Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

XiaoBei Image to VBA · xiao24bei | DecimalAI

Codingby xiao24bei

XiaoBei Image to VBA

Use when users want XiaoBei skill / xiaobei-skill / 小北在读研 style academic image-to-VBA reconstruction: convert academic figures, scientific diagrams, slides, screenshots, or other images into VBA, Office drawing code, PowerPoint/Excel/Word shapes, editable shape reconstruction, 1:1 recreation, pixel-like approximation, or hybrid editable Office shape reconstruction.

Never made a case worseMaintained

SkillScore

SkillScore77/100

Test results 77

Safety

3/3 safety checks passed

Code scan
Intent
Content

What it changes

45%77%of test cases passed +32 pts

Without skill10 of 22

With skill17 of 22

7 fixed10 passed either way5 still failing0 broke

Measured on gemini-3.6-flash9/29/2026v121 comparable

Track record

Agent runs: 0; All time

See one task, with and without the skill

One test taskReconstruct a visual diagram image into a PowerPoint VBA macro. Provide a complete element breakdown table, analysis report, self-check summary, and code comments. Unless I ask for English, what language should be used for the breakdown, reports, and code comments?

Without the skillMissed

With the skillPassed

Same model, same task. Answers are capped and redacted.

Use this skill

Your agent runs on

Adds every file the skill has, pinned so you can update it. No account needed.

pip install "decimalai>=0.15"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba
claude --plugin-dir "$(git rev-parse --show-toplevel 2>/dev/null || pwd)/.agents"

The same install from Python, for a notebook or a setup script.

Install

pip install "decimalai>=0.15"

Add the skill

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

print("Added. Start Claude Code with the command below.")

Start Claude Code with it

claude --plugin-dir "$(git rev-parse --show-toplevel 2>/dev/null || pwd)/.agents"

Type these in a Claude Code session. No account or API key. It writes the skill and any files it bundles, and asks first when its safety rating isn't Passed.

Add the DecimalAI plugin, once

/plugin marketplace add decimal-labs/decimalai-claude-plugins
/plugin install decimalai@decimalai

Install this skill

/decimalai:install xiao24bei-xiaobei-skill-image-to-vba

The plugin needs python3 and curl on your machine.

No Python or account. Claude Code picks it up in its next session.

Save it into your project

curl -fsSL --create-dirs -o .claude/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md https://app.decimal.ai/s/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Copy SKILL.md

Save it in your project, as

.claude/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Claude Code docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[claude-agent-sdk]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

import asyncio

from claude_agent_sdk import ClaudeAgentOptions
from decimalai.claude_agent_sdk import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

async def main():
    question = "Hello!"
    async with support.run(question) as run:  # one user message, one run
        async for message in run.query(options=ClaudeAgentOptions(max_turns=6)):
            print(message)

# In a notebook or async code: await main()
asyncio.run(main())

Prefer its own loader? ClaudeAgentOptions(plugins=[skills_plugin()], skills="all") reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

The same install from Python, for a notebook or a setup script. agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[claude-agent-sdk]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

import asyncio

from claude_agent_sdk import ClaudeAgentOptions
from decimalai.claude_agent_sdk import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

async def main():
    question = "Hello!"
    async with support.run(question) as run:  # one user message, one run
        async for message in run.query(options=ClaudeAgentOptions(max_turns=6)):
            print(message)

# In a notebook or async code: await main()
asyncio.run(main())

Prefer its own loader? ClaudeAgentOptions(plugins=[skills_plugin()], skills="all") reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from claude_agent_sdk import ClaudeAgentOptions, query

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
options = ClaudeAgentOptions(system_prompt=f"You are a helpful agent.\n\n{skill}", max_turns=6)

async def main():
    async for message in query(prompt="Hello!", options=options):
        print(message)

# In a notebook or async code: await main()
asyncio.run(main())

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Claude Agent SDK docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[openai-agents]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

from agents import Agent, Runner
from decimalai.openai_agents import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    name=support.name,
    instructions=support.instructions,  # your prompt, then the skill list
    tools=[*support.tools],             # your tools, then load_skill
)
# In a notebook or async code: result = await Runner.run(agent, "Hello!")
result = Runner.run_sync(agent, "Hello!")

Using SandboxAgent? capabilities=[Shell(), decimalai.openai_agents.sandbox_skills()] points its own loader at the same folder.

Install

pip install "decimalai[openai-agents]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

from agents import Agent, Runner
from decimalai.openai_agents import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    name=support.name,
    instructions=support.instructions,  # your prompt, then the skill list
    tools=[*support.tools],             # your tools, then load_skill
)
# In a notebook or async code: result = await Runner.run(agent, "Hello!")
result = Runner.run_sync(agent, "Hello!")

Using SandboxAgent? capabilities=[Shell(), decimalai.openai_agents.sandbox_skills()] points its own loader at the same folder.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from agents import Agent, Runner

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = Agent(name="support", instructions=f"You are a helpful agent.\n\n{skill}")
# In a notebook or async code: result = await Runner.run(agent, "Hello!")
result = Runner.run_sync(agent, "Hello!")

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

OpenAI Agents docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[openai]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

import json

from openai import OpenAI
from decimalai.openai import agent_config

client = OpenAI()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your tools, as Chat Completions takes them, and what answers each.
TOOLS = [{"type": "function", "function": {
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters": {"type": "object", "properties": {"order_id": {"type": "string"}},
                   "required": ["order_id"]},
}}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    # run.system is your prompt, then the skill list.
    messages = [{"role": "system", "content": run.system}, run.message]
    while True:
        reply = client.chat.completions.create(
            model="gpt-5.4-mini",
            messages=messages,
            tools=[*TOOLS, *run.tools],  # your tools, then load_skill
        ).choices[0].message
        messages.append(reply)
        if not reply.tool_calls:
            break
        for call in reply.tool_calls:
            # load_skill is answered by the run; your own tools by FUNCTIONS.
            function = FUNCTIONS.get(call.function.name)
            messages.append(run.tool_result(call) or {
                "role": "tool", "tool_call_id": call.id,
                "content": function(**json.loads(call.function.arguments)) if function
                else f"There is no tool named {call.function.name!r}.",
            })
print(reply.content)

# The Responses API takes run.response_tools instead of run.tools.

Install

pip install "decimalai[openai]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

import json

from openai import OpenAI
from decimalai.openai import agent_config

client = OpenAI()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your tools, as Chat Completions takes them, and what answers each.
TOOLS = [{"type": "function", "function": {
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters": {"type": "object", "properties": {"order_id": {"type": "string"}},
                   "required": ["order_id"]},
}}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    # run.system is your prompt, then the skill list.
    messages = [{"role": "system", "content": run.system}, run.message]
    while True:
        reply = client.chat.completions.create(
            model="gpt-5.4-mini",
            messages=messages,
            tools=[*TOOLS, *run.tools],  # your tools, then load_skill
        ).choices[0].message
        messages.append(reply)
        if not reply.tool_calls:
            break
        for call in reply.tool_calls:
            # load_skill is answered by the run; your own tools by FUNCTIONS.
            function = FUNCTIONS.get(call.function.name)
            messages.append(run.tool_result(call) or {
                "role": "tool", "tool_call_id": call.id,
                "content": function(**json.loads(call.function.arguments)) if function
                else f"There is no tool named {call.function.name!r}.",
            })
print(reply.content)

# The Responses API takes run.response_tools instead of run.tools.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from openai import OpenAI

client = OpenAI()
skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
reply = client.chat.completions.create(
    model="gpt-5.4-mini",
    messages=[
        {"role": "system", "content": f"You are a helpful agent.\n\n{skill}"},
        {"role": "user", "content": "Hello!"},
    ],
).choices[0].message

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

OpenAI SDK docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[anthropic]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

import anthropic
from decimalai.anthropic import agent_config

client = anthropic.Anthropic()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

question = "Hello!"
with support.run(question) as run:  # one user message, one run
    answer = client.beta.messages.tool_runner(
        model="claude-haiku-5-5",
        max_tokens=4096,
        system=run.system,     # your prompt, then the skill list
        tools=[*run.tools],    # your tools, then load_skill
        messages=[{"role": "user", "content": question}],
    ).until_done()

Install

pip install "decimalai[anthropic]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

import anthropic
from decimalai.anthropic import agent_config

client = anthropic.Anthropic()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

question = "Hello!"
with support.run(question) as run:  # one user message, one run
    answer = client.beta.messages.tool_runner(
        model="claude-haiku-5-5",
        max_tokens=4096,
        system=run.system,     # your prompt, then the skill list
        tools=[*run.tools],    # your tools, then load_skill
        messages=[{"role": "user", "content": question}],
    ).until_done()

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
import anthropic

client = anthropic.Anthropic()
skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
reply = client.messages.create(
    model="claude-haiku-5-5",
    max_tokens=4096,
    system=f"You are a helpful agent.\n\n{skill}",
    messages=[{"role": "user", "content": "Hello!"}],
)

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Anthropic docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[google-genai]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

from google import genai
from google.genai import types
from decimalai.google_genai import agent_config

client = genai.Client()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your functions, as declarations for the model, and what answers each.
TOOLS = [{"function_declarations": [{
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters_json_schema": {"type": "object", "properties": {"order_id": {"type": "string"}},
                               "required": ["order_id"]},
}]}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    config = types.GenerateContentConfig(
        system_instruction=run.system,     # your prompt, then the skill list
        tools=[*TOOLS, *run.tool_params],  # your functions, then load_skill
    )
    contents = [{"role": "user", "parts": [{"text": run.message}]}]
    while True:
        response = client.models.generate_content(
            model="gemini-3.5-flash", contents=contents, config=config,
        )
        if not response.function_calls:
            break
        contents.append(response.candidates[0].content)
        parts = []
        for call in response.function_calls:
            # load_skill is answered by the run; your own functions by FUNCTIONS.
            # A loop of your own rather than automatic function calling, which
            # raises KeyError when the model asks for a function you did not pass.
            function = FUNCTIONS.get(call.name)
            parts.append(run.tool_result(call) or {"function_response": {
                "name": call.name, "id": call.id,
                "response": {"result": function(**(call.args or {})) if function
                             else f"There is no function named {call.name!r}."},
            }})
        contents.append({"role": "user", "parts": parts})
print(response.text)

Install

pip install "decimalai[google-genai]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

from google import genai
from google.genai import types
from decimalai.google_genai import agent_config

client = genai.Client()
# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")


def lookup_order(order_id: str) -> str:
    return f"Order {order_id} shipped yesterday."


# Your functions, as declarations for the model, and what answers each.
TOOLS = [{"function_declarations": [{
    "name": "lookup_order",
    "description": "Look up the status of an order.",
    "parameters_json_schema": {"type": "object", "properties": {"order_id": {"type": "string"}},
                               "required": ["order_id"]},
}]}]
FUNCTIONS = {"lookup_order": lookup_order}

question = "Where is order A1001?"
with support.run(question) as run:  # one user message, one run
    config = types.GenerateContentConfig(
        system_instruction=run.system,     # your prompt, then the skill list
        tools=[*TOOLS, *run.tool_params],  # your functions, then load_skill
    )
    contents = [{"role": "user", "parts": [{"text": run.message}]}]
    while True:
        response = client.models.generate_content(
            model="gemini-3.5-flash", contents=contents, config=config,
        )
        if not response.function_calls:
            break
        contents.append(response.candidates[0].content)
        parts = []
        for call in response.function_calls:
            # load_skill is answered by the run; your own functions by FUNCTIONS.
            # A loop of your own rather than automatic function calling, which
            # raises KeyError when the model asks for a function you did not pass.
            function = FUNCTIONS.get(call.name)
            parts.append(run.tool_result(call) or {"function_response": {
                "name": call.name, "id": call.id,
                "response": {"result": function(**(call.args or {})) if function
                             else f"There is no function named {call.name!r}."},
            }})
        contents.append({"role": "user", "parts": parts})
print(response.text)

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from google import genai
from google.genai import types

client = genai.Client()
skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
response = client.models.generate_content(
    model="gemini-3.5-flash",
    contents="Hello!",
    config=types.GenerateContentConfig(system_instruction=f"You are a helpful agent.\n\n{skill}"),
)

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Gemini API docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[langchain]" langchain-openai
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

from langchain.agents import create_agent
from decimalai.langchain import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = create_agent(
    "openai:gpt-5.4-mini",
    tools=[],                         # your tools; the middleware adds load_skill
    middleware=[support.middleware],  # your prompt, the skill list and load_skill
)
result = agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

Install

pip install "decimalai[langchain]" langchain-openai

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

from langchain.agents import create_agent
from decimalai.langchain import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = create_agent(
    "openai:gpt-5.4-mini",
    tools=[],                         # your tools; the middleware adds load_skill
    middleware=[support.middleware],  # your prompt, the skill list and load_skill
)
result = agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from langchain.agents import create_agent

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = create_agent("openai:gpt-5.4-mini", tools=[], system_prompt=f"You are a helpful agent.\n\n{skill}")
result = agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

LangChain docs

CrewAI's own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" "crewai>=1.15.27"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

import decimalai.crewai
from crewai import Agent, Crew, Task

agent = Agent(
    role="Support", goal="Help the customer", backstory="A support agent.",
    llm=llm,                               # your own CrewAI LLM
    skills=decimalai.crewai.skills(),      # the installed folder
    tools=decimalai.crewai.skill_tools(),  # reads a skill's files, runs its scripts
)
task = Task(description="Hello!", expected_output="an answer", agent=agent)
Crew(agents=[agent], tasks=[task]).kickoff()

The same install from Python, for a notebook or a setup script. CrewAI's own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" "crewai>=1.15.27"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

import decimalai.crewai
from crewai import Agent, Crew, Task

agent = Agent(
    role="Support", goal="Help the customer", backstory="A support agent.",
    llm=llm,                               # your own CrewAI LLM
    skills=decimalai.crewai.skills(),      # the installed folder
    tools=decimalai.crewai.skill_tools(),  # reads a skill's files, runs its scripts
)
task = Task(description="Hello!", expected_output="an answer", agent=agent)
Crew(agents=[agent], tasks=[task]).kickoff()

No DecimalAI package. CrewAI's own loader reads the folder, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from crewai import Agent, Crew, Task

agent = Agent(
    role="Support", goal="Help the customer", backstory="A support agent.",
    llm=llm,                         # your own CrewAI LLM
    skills=[Path(".agents/skills")],
)
task = Task(description="Hello!", expected_output="an answer", agent=agent)
Crew(agents=[agent], tasks=[task]).kickoff()

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

CrewAI docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[adk]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

from decimalai.adk import DecimalAIPlugin, agent_config
from google.adk.agents import LlmAgent
from google.adk.apps import App

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = LlmAgent(
    name="support",          # ADK needs a Python identifier
    model="gemini-3.5-flash",
    instruction=support.instruction,  # your prompt, then the skill list
    tools=[*support.tools],           # your tools, then load_skill
)
# DecimalAIPlugin makes each invocation of runners built from this App one trace.
app = App(name="support", root_agent=agent, plugins=[DecimalAIPlugin()])

Prefer ADK's own loader? SkillToolset(skills=decimalai.adk.skills(), ...) reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

Install

pip install "decimalai[adk]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

from decimalai.adk import DecimalAIPlugin, agent_config
from google.adk.agents import LlmAgent
from google.adk.apps import App

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = LlmAgent(
    name="support",          # ADK needs a Python identifier
    model="gemini-3.5-flash",
    instruction=support.instruction,  # your prompt, then the skill list
    tools=[*support.tools],           # your tools, then load_skill
)
# DecimalAIPlugin makes each invocation of runners built from this App one trace.
app = App(name="support", root_agent=agent, plugins=[DecimalAIPlugin()])

Prefer ADK's own loader? SkillToolset(skills=decimalai.adk.skills(), ...) reads the same folder. Use one or the other: a skill delivered both ways reaches the prompt twice.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from google.adk.agents import LlmAgent

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = LlmAgent(
    name="support",
    model="gemini-3.5-flash",
    instruction=lambda context: f"You are a helpful agent.\n\n{skill}",
)

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Google ADK docs

agent_config() hands your agent the skill. With no DecimalAI key it reads your project's files and sends nothing. Needs SDK 0.27 or newer.

Install

pip install "decimalai[pydantic-ai]"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

from pydantic_ai import Agent
from decimalai.pydantic_ai import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    "openai:gpt-4o-mini",
    name=support.name,
    tools=[],                 # your tools; the capability adds load_skill
    capabilities=[support],   # your prompt, the skill list and load_skill
)
# In a notebook or async code: result = await agent.run("Hello!")
result = agent.run_sync("Hello!")

On the Pydantic AI harness, decimalai.pydantic_ai_harness.skills(project) uses its own loader instead.

Install

pip install "decimalai[pydantic-ai]"

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

from pydantic_ai import Agent
from decimalai.pydantic_ai import agent_config

# With no DECIMAL_API_KEY this reads the skills your project lists, from disk.
support = agent_config("support", fallback="You are a helpful agent.")

agent = Agent(
    "openai:gpt-4o-mini",
    name=support.name,
    tools=[],                 # your tools; the capability adds load_skill
    capabilities=[support],   # your prompt, the skill list and load_skill
)
# In a notebook or async code: result = await agent.run("Hello!")
result = agent.run_sync("Hello!")

On the Pydantic AI harness, decimalai.pydantic_ai_harness.skills(project) uses its own loader instead.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from pydantic_ai import Agent

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = Agent("openai:gpt-4o-mini", instructions=f"You are a helpful agent.\n\n{skill}")
# In a notebook or async code: result = await agent.run("Hello!")
result = agent.run_sync("Hello!")

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Pydantic AI docs

Deep Agents' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" deepagents
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

import decimalai.deepagents
from deepagents import create_deep_agent

agent = create_deep_agent(model=model, **decimalai.deepagents.skills())
agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

The default backend reads files but has no shell. To run a skill's scripts, pass skills(backend=LocalShellBackend(...)).

The same install from Python, for a notebook or a setup script. Deep Agents' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" deepagents

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

import decimalai.deepagents
from deepagents import create_deep_agent

agent = create_deep_agent(model=model, **decimalai.deepagents.skills())
agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

The default backend reads files but has no shell. To run a skill's scripts, pass skills(backend=LocalShellBackend(...)).

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from deepagents import create_deep_agent

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = create_deep_agent(model=model, system_prompt=f"You are a helpful agent.\n\n{skill}")
agent.invoke({"messages": [{"role": "user", "content": "Hello!"}]})

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Deep Agents docs

Its own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" agent-framework
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

import asyncio

from agent_framework import Agent
from decimalai.agent_framework import skills

agent = Agent(client=client, instructions="You are a helpful agent.", context_providers=[skills()])
asyncio.run(agent.run("Hello!"))

It caches skills by default, so restart the process after decimalai skills update.

The same install from Python, for a notebook or a setup script. Its own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" agent-framework

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

import asyncio

from agent_framework import Agent
from decimalai.agent_framework import skills

agent = Agent(client=client, instructions="You are a helpful agent.", context_providers=[skills()])
asyncio.run(agent.run("Hello!"))

It caches skills by default, so restart the process after decimalai skills update.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from agent_framework import Agent

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = Agent(client=client, instructions=f"You are a helpful agent.\n\n{skill}")
asyncio.run(agent.run("Hello!"))

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Microsoft Agent Framework docs

Strands' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" strands-agents
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

Add it to your agent

from decimalai.strands import skills
from strands import Agent
from strands.vended_tools import file_editor, shell

agent = Agent(
    model=model,                # always pass one: the default is Amazon Bedrock
    system_prompt="You are a helpful agent.",
    tools=[file_editor, shell],
    plugins=[skills()],
)
agent("Hello!")

Strands loads an agent's skills once, so build the agent again after decimalai skills update.

The same install from Python, for a notebook or a setup script. Strands' own skill loader reads the installed folder. No DecimalAI account needed.

Install

pip install "decimalai>=0.16" strands-agents

Add it to your agent

from decimalai import skillpkg

# Writes .agents/skills and .decimal/skills.toml. Safe to run twice.
result = skillpkg.add("xiao24bei-xiaobei-skill-image-to-vba")
if result.exit_code:
    raise SystemExit(result.problems)

from decimalai.strands import skills
from strands import Agent
from strands.vended_tools import file_editor, shell

agent = Agent(
    model=model,                # always pass one: the default is Amazon Bedrock
    system_prompt="You are a helpful agent.",
    tools=[file_editor, shell],
    plugins=[skills()],
)
agent("Hello!")

Strands loads an agent's skills once, so build the agent again after decimalai skills update.

No DecimalAI package. The whole skill goes in every prompt, and nothing pins or updates it.

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Give it to your agent

from pathlib import Path
from strands import Agent

skill = Path(".agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md").read_text()
agent = Agent(model=model, system_prompt=f"You are a helpful agent.\n\n{skill}")
agent("Hello!")

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Strands docs

Adds every file the skill has, pinned so you can update it. Cursor, Copilot, Cline and Warp read this folder by themselves.

pip install "decimalai>=0.15"
decimalai skills add xiao24bei-xiaobei-skill-image-to-vba

No Python or account. Cursor, Copilot, Cline and Warp read this folder by themselves.

Save it into your project

curl -fsSL --create-dirs -o .agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md https://app.decimal.ai/s/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Copy SKILL.md

Save it in your project, as

.agents/skills/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Cursor and other editors docs

Fetch the file and give its text to your model as instructions. No account or package needed.

Get the file

curl -fsSL https://app.decimal.ai/s/xiao24bei-xiaobei-skill-image-to-vba/SKILL.md

Read the summary as JSON

curl -fsSL https://app.decimal.ai/s/xiao24bei-xiaobei-skill-image-to-vba.json

This gets SKILL.md only. The skill's 12 other files are not included.

Other ways to add it

Other setups docs

You don't need a DecimalAI account.

Version: v1, tested 9/29/2026
Source: xiao24bei/xiaobei-skill ↗
License: Apache-2.0

Manage it in DecimalAI

Saved prompts and skills, pinned versions, and traces of this skill in your runs. DecimalAI does not run your agent.

Claude Code reads files from your project, so DecimalAI config does not apply there.

On Claude Agent SDK: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On OpenAI Agents: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On OpenAI SDK: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Anthropic: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Gemini API: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On LangChain: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On CrewAI: set DECIMAL_API_KEY and add decimalai.init(crewai=True) to record runs and skill use. The skill version stays in your repo.

On Google ADK: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Pydantic AI: save the skill to an agent here, set DECIMAL_API_KEY, and the same code uses it with no command to run. Changes you save reach the next run with no redeploy.

On Deep Agents: set DECIMAL_API_KEY and add decimalai.init(langchain=True) to record runs and skill use. The skill version stays in your repo.

On Microsoft Agent Framework: set DECIMAL_API_KEY and add decimalai.init(otel=True, agent_name="support") to record runs and skill use. The skill version stays in your repo.

On Strands: set DECIMAL_API_KEY and add decimalai.init(otel=True, agent_name="support") to record runs and skill use. The skill version stays in your repo.

Editors read files from your project, so DecimalAI config does not apply there.

Anywhere else: record a use with decimalai.log_skill_activation() inside a trace.

Use this skill

SKILL.md

XiaoBei Image to VBA

小北在读研出品的 academic image-to-vba reconstruction skill. The workflow focuses on editable Office Shapes first, with carefully labeled raster crops only when complex scientific artwork should be preserved.

Output Language

All analysis, manifest tables, self-check reports, run reports, and conversational responses produced by this skill MUST be written in Simplified Chinese (简体中文). This includes:

视觉拆解、元素清单、坐标说明、自检结论、运行报告、给用户的最终回复。
VBA 代码内的注释也使用中文（例如 ' 标题区域背景），便于用户后续维护。

仅以下内容保留英文原样，不要翻译：

VBA 关键字、API 名、属性名（Sub、Shapes.AddShape、Fill.ForeColor.RGB 等）。
形状名前缀 AITVBA_ 与脚本路径、文件名。
颜色十六进制值与 RGB(...) 调用本身。

如果用户在对话里明确要求换成英文或其它语言，则按用户指定的语言输出，但 manifest 字段名仍保留英文 key 以便机器解析。

Purpose

Convert a reference image into a complete, runnable Office VBA macro that reconstructs the visual using editable Shapes, TextBoxes, Lines, Freeforms, fills, gradients, transparency, shadows, and ZOrder. When the image contains complex, distinctive graphic elements that would become noisy or low-value if rebuilt as hundreds of shapes, preserve those elements as cropped local image assets and reconstruct the surrounding structure with editable VBA Shapes. The core loop is: image -> environment detection -> visual decomposition -> editable-vs-preserved element plan -> Office coordinate model -> VBA Shapes code and cropped assets -> run or prepare the macro in the available presentation app -> self-check -> iterative correction.

Prefer editable Office Shapes for structure, labels, arrows, boxes, chart scaffolds, and readable text. Preserve complex graphic elements only as local cropped assets when that improves fidelity and keeps the VBA maintainable. Do not embed the full source image as a shortcut unless the user explicitly permits it or chooses the background-image-assisted option for complex photos. A PNG/JPEG/PDF preview is never the final deliverable for this skill; it is only allowed as a diagnostic screenshot after an editable Office reconstruction has been produced or attempted.

Default Deliverable Contract

When the user provides an image and asks for reconstruction, the default target is a presentation deck unless another Office host is explicit. Prefer Microsoft PowerPoint automation when available, but support WPS users and no-local-office users with a truthful fallback path. The agent must produce:

A complete VBA macro file or code block that reconstructs the image with editable Office Shapes.
Any preserved local image assets needed for hybrid reconstruction, with clear labels showing they are not editable Shapes.
A presentation file created from that macro when local automation is available, or a clear statement that local presentation automation was blocked or unsupported.
The generated deck opened for the user when local PowerPoint/WPS opening is available.
A short run report that says whether the macro was executed, whether the deck opened, and where the generated .bas / .pptx / .pptm files are located.

Do not satisfy an image-to-VBA request by generating only a PNG image, raster preview, static screenshot, or visual mockup. If a preview image is useful, create it only in addition to the editable VBA/PPT deliverables.

Hybrid Preservation Mode

Use hybrid preservation when any of the hard triggers below fire. Do not rely on subjective judgment about whether the result "looks cluttered" — model self-assessment is biased toward editable reconstruction, which causes complex artwork to be redrawn poorly. The goal is to preserve high-complexity artwork while keeping the diagram's editable structure editable.

Hard preservation triggers (hit any one → the element MUST be preserved as a raster crop):

The element is a microscopy image, 3D render, real photograph, or photo-realistic illustration.
The element contains photo-grade gradients, textures, noise, or organic shading that cannot be matched by a few solid fills or simple gradients.
The element is a logo, trademarked mark, or branded artwork.
A faithful editable reconstruction would require more than ~15 Shapes for that single element.
The element is hand-drawn, painterly, or has stroke-width variation that cannot be matched by simple Line/Freeform.
The user explicitly named the element as "keep as image", "保留", "不拆", or equivalent.

If none of the triggers fire, treat the element as editable. If a trigger fires but the user demanded "all editable", explain the fidelity tradeoff first and get permission before redrawing.

Classify every visible element into one of three buckets:

Editable Shapes: text, arrows, boxes, connectors, axes, callouts, tables, UI panels, simple icons, repeated markers, and diagram scaffolds.
Preserved raster elements: distinctive illustrations, logos, scientific renderings, microscopy/photo crops, 3D objects, complex texture patches, dense molecular/mesh graphics, screenshots inside device frames, and hand-drawn/artistic elements whose shape-by-shape reconstruction would be noisy.
Background/base: plain solid/gradient backgrounds should be rebuilt as shapes; full-image background insertion is allowed only when the user explicitly asks for it or the source is photo-like.

Hybrid rules:

Preserve only the smallest useful crop around each complex element; do not insert the entire source image just to preserve a small object.
Put preserved assets behind or between editable layers according to the original z-order.
Recreate nearby labels, arrows, boxes, highlights, and callouts as editable VBA objects even when they overlap preserved artwork.
Name preserved image shapes with the generated prefix, for example AITVBA_Raster_Microscope_01.
Save cropped assets next to the generated VBA/deck, preferably under an assets/ subdirectory.
In the final response, list preserved elements separately from editable elements so the user knows which parts remain raster.
If the user requests "all editable", do not use preserved raster elements unless you first explain the fidelity tradeoff and get permission.

保留元素的几何契约 (Preservation Geometry Contract)

保留元素出现"裁错 / 变形 / 缩放比例不对"是 hybrid 模式最常见的失败。下面这套契约是硬性的,违反任何一条都视为本轮 render-verify 不通过,必须回到 manifest 修正。

裁剪 bbox 必须等于 manifest bbox_px,不准 padding,不准外扩,不准内缩
- 调用 scripts/crop_preserved_elements.py 时,传入的坐标必须与 manifest 中该元素的 bbox_px 完全一致(整数像素)。
- 不要为了"美观"或"留出阴影"自行扩边。如果原图边界外有阴影/光晕需要保留,那应在 manifest 里就把 bbox_px 扩到包含阴影的范围,而不是裁的时候临时加。
- 裁出来的 PNG 必须保留原图 alpha 通道(若有),不要预先合成到白底。
插入时必须保持裁剪图自身的宽高比
- 在 VBA 里写 Shapes.AddPicture 后,只能用 ratio-safe 缩放:已知裁剪图原始像素宽高 cw_px, ch_px,目标插入区域宽高 tw_pt, th_pt,则:
  - s = min(tw_pt / cw_px, ch_px > 0 ? th_pt / ch_px : 1)
  - 实际插入的形状宽 = cw_px * s,高 = ch_px * s
- 严禁直接用 Width = X : Height = Y 把图片强行拉到 manifest 的目标尺寸,除非目标尺寸恰好与裁剪图等比。一旦目标 bbox 与裁剪源比例不一致,要么调整目标 bbox 使之与源等比,要么在目标区域内居中留白。
- 不允许给 AddPicture 之后再设置 LockAspectRatio = msoFalse 然后独立改宽高。
裁剪图与目标 bbox 的比例必须一致
- manifest 中保留元素的 bbox_px 既描述源图坐标,也决定目标 bbox 的宽高比。
- 如果裁出来的 PNG 实际宽高比与 manifest 的 bbox_px 比例偏差超过 ±2%,说明裁剪坐标错了,必须重新裁,而不是在插入时拉伸去"对齐"。
不允许旋转 / 翻转 / 任意 freeform 蒙版
- 默认 Rotation = 0,FlipH = msoFalse,FlipV = msoFalse。除非原图本来就是斜的,且 manifest 里该行显式给出了 rotation_deg 字段,否则不许动。
- 不要给保留图片套圆角矩形蒙版、椭圆蒙版、freeform 蒙版,这会改变它的视觉形状。需要圆角的,只能在 manifest 里把元素本身标为 editable,而不是把矩形裁剪强行变形。
插入位置使用 ratio-safe 坐标转换
- 用第 3 节的 scale / offset_x_pt / offset_y_pt 公式把 bbox_px 的 (x, y) 转成插入点 (Left, Top)。
- 不要用独立 X/Y 拉伸的版本,即使画布与原图同比也优先用 ratio-safe 形式,避免一处偷懒导致其它元素错位。
裁剪资产的命名与可追溯
- 文件名 = assets/<id>_<语义短名>.png,例如 assets/R03_microscope_left.png。
- 在 manifest 中保留该文件的 SHA-1 前 8 位或文件大小,render-verify 闭环里若某个保留元素差异大,先核对这一行哈希是否变了,排除"裁了又被覆盖"的情况。

保留元素自检 (额外的第 3.5 项 self-check)

在第 6 节 Mandatory Self-Checks 的"Preservation check"基础上,每一轮 render-verify 都额外运行下面四问,任何一问回答不出"是",该元素就不算合格:

这张裁剪图的宽高比 (cw_px / ch_px) 与 manifest bbox_px 的宽高比之差 < 2% 吗?
VBA 中该 AddPicture 之后的 Shape.Width / Shape.Height 比值与裁剪图原始比值之差 < 2% 吗?
该形状的 Rotation/FlipH/FlipV 是否与 manifest 显式声明一致(默认 0/false/false)?
渲染图中该形状的边界与原图中对应区域的边界,在 ratio-safe 映射后误差是否 < 画布短边的 2%?

不达标时,优先调整 manifest 的 bbox_px 与 crop_path,而不是在 VBA 里硬拉尺寸。

Environment Compatibility

Before trying to materialize a deck, identify the user's local runtime. Use scripts/detect_office_environment.py when local shell access is available; otherwise infer from the user's stated OS and installed office suite.

Choose the delivery path in this order:

Windows + Microsoft PowerPoint: generate .bas, run scripts/vba_lint.py, then try scripts/run_powerpoint_vba_windows.ps1. Report if PowerPoint Trust Center blocks VBA import.
macOS + Microsoft PowerPoint: generate .bas, run scripts/vba_lint.py, then try scripts/run_powerpoint_vba_macos.py. Report if AppleScript automation or Office macro security blocks execution.
WPS Presentation / WPS Office only: generate conservative PowerPoint-compatible VBA and a .bas file. Do not assume automatic VBA import/run is available. If WPS can be opened locally, open it and provide WPS-specific manual import/run steps. If WPS macro support is absent or unverified, say that plainly.
No local presentation app or unsupported platform: still generate the .bas VBA source. When feasible, also generate an editable .pptx fallback from the same shape plan, clearly labeled as a non-VBA fallback. Do not substitute a PNG as the final output.

Use WPS-compatible mode when WPS is the only available app:

Prefer basic Shapes, Lines, TextBoxes, solid fills, RGB colors, simple transparency, and simple shadows.
Avoid depending on advanced PowerPoint-only effects, complex TextFrame2 behavior, gradient stops, and macro security settings unless verified locally.
Save outputs in Microsoft-compatible formats (.bas, .pptx, .pptm) rather than WPS-only formats unless the user asks otherwise.
Label WPS results as compatibility-targeted, not fully PowerPoint-verified, unless they were actually opened and run in WPS.

Workflow

1. Input Recognition

Identify the target Office host:

Use PowerPoint when the user does not specify a host.
Use Excel when the user asks for a worksheet, dashboard, spreadsheet, chart sheet, grid, or Excel Shapes.
Use Word when the user asks for a document page, report, Word canvas, or inline/anchored drawing.

Ask a concise clarification only when the target host materially changes the solution and cannot be inferred. Otherwise proceed with the default presentation target, detect the available runtime, and attempt the best supported create/open path.

Classify the source image:

Vector/flat graphic: icons, diagrams, simple logos, flowcharts, infographics. Aim for high-fidelity editable shape reconstruction.
UI screenshot/poster: cards, bars, text, panels, tables, app screens. Aim for detailed reconstruction with editable text and shape groups.
Mixed scientific/academic figure with complex artwork plus labels: default to hybrid preservation for the complex artwork and editable reconstruction for text, arrows, boxes, connectors, legends, and callouts.
Complex photo/texture/person: explain limits. Offer:
- A. Pure VBA Shapes approximate reconstruction.
- B. Hybrid preservation of local crops plus editable VBA overlays.
- C. Use the image as a full background and add editable VBA overlays.

2. Image Parsing

Inspect the image before coding. Record:

Canvas ratio, orientation, approximate pixel dimensions, and target Office canvas size.
Background: solid, gradient, image-like, texture, transparency, page margins.
Major objects: shapes, panels, icons, chart elements, tables, arrows, connectors, callouts.
Reconstruction bucket for each major object: editable shape, preserved raster crop, or background/base.
Preserved raster crop coordinates and asset names when hybrid mode is used.
Text: exact words when legible; use placeholder text with a comment when uncertain.
Colors: approximate RGB values; keep a palette list for reuse.
Lines: stroke width, caps, joins, dash style, arrowheads.
Effects: shadows, blur-like approximations, gradients, transparency, glow, bevel-like effects.
Layers: back-to-front order and overlaps.
Alignment: margins, gutters, repeated spacing, baseline grids, visual centers.

For detailed pattern examples, load references/vba-shape-patterns.md only when needed.

2.5 Element Manifest (mandatory)

Before writing any VBA, produce an explicit element manifest. This step is mandatory — do not skip to coordinate modeling or code generation without it. Skipping the manifest is the root cause of three frequent failures: missed preservation, drifted layout, and arrows pointing to wrong endpoints.

Output the manifest as a Markdown table or JSON list with these fields per element:

id: stable short id, for example E01, R01, L01. Use E* for editable, R* for preserved raster, L* for lines/arrows/connectors, B* for background.
type: rect, ellipse, textbox, arrow, line, freeform, group, raster_crop, background, etc.
bucket: editable, preserved, or background. Apply the hard preservation triggers from "Hybrid Preservation Mode" — do not guess.
bbox_px: (x, y, w, h) in source-image pixels. Required for every element except pure connectors. Estimate visually but commit to numbers; do not write ? or "approx".
style: fill color (hex or RGB), stroke color, stroke width, font size, font weight, opacity, shadow, gradient endpoints, etc. Use null only when the element has no such property.
text: exact text content for textboxes; null otherwise. Mark uncertain readings with a trailing (?).
endpoints: for arrows/lines/connectors only. Mandatory and strict. Format must be from:(x1_px,y1_px) -> to:(x2_px,y2_px) [anchor:<id>.<side>=<id>.<side>]. Both absolute pixel coordinates AND the semantic anchor (which element's which side) are required when the line connects two elements. Reasons:
- 仅靠"靠近 A"或"指向 B"会导致箭头乱指 — 模型会按印象画。
- 必须在原图上目测出起点和终点的像素坐标，并写明粘连到哪个元素的哪条边（top/right/bottom/left/center）。
- 如果箭头中间有折线（elbow / orthogonal），用 via:(x,y);(x,y) 字段补充中间拐点像素坐标。
- 如果起点或终点是悬空的（不连接任何元素），写 anchor:free。
- 生成 VBA 时，连接两元素的箭头必须用 Shapes.AddConnector 并 BeginConnect / EndConnect 到具体形状，而不是用绝对坐标画一根 AddLine。这样箭头会随被连接元素位置自动跟随,不会乱指。
z_order: integer, lower draws first.
preserve_reason: required when bucket = preserved. State which hard trigger fired.
crop_path: required when bucket = preserved. Use assets/<id>_<short_name>.png.

Example minimal row:

| id  | type      | bucket    | bbox_px            | style                       | text       | endpoints                    | z | preserve_reason | crop_path                  |
|-----|-----------|-----------|--------------------|-----------------------------|------------|------------------------------|---|-----------------|----------------------------|
| E01 | rect      | editable  | 120,80,200,60      | fill=#FF8800,stroke=#333,2pt| "Input"    | —                            | 2 | —               | —                          |
| R01 | raster    | preserved | 400,200,180,180    | —                           | —          | —                            | 1 | microscopy image| assets/R01_microscopy.png  |
| L01 | arrow     | editable  | —                  | stroke=#000,2pt,arrow_end   | —          | from:E01.right -> to:R01.left| 3 | —               | —                          |

Manifest gate: do not generate VBA until every visible element appears in the manifest. The number of rows in the manifest is the lower bound on the number of Shapes the macro will create (preserved elements count as one AddPicture each). If a region of the image has no manifest row, you have not parsed that region — go back to step 2.

3. Coordinate Modeling

Model the image in Office points by default.

Preserve the original aspect ratio.
For PowerPoint, prefer a proportional custom slide size based on the image ratio. Use 960 pt as the long side unless the user requests a specific slide size.
If the user requires a standard slide size such as 16:9, fit the image uniformly within that canvas and add offset_x_pt / offset_y_pt; do not stretch the image independently on X and Y.
Only use independent X/Y scaling when the target canvas exactly matches the source aspect ratio or when the user explicitly accepts distortion.
Convert each image coordinate using the ratio-safe method:
- scale = min(target_width_pt / image_width_px, target_height_pt / image_height_px)
- offset_x_pt = (target_width_pt - image_width_px * scale) / 2
- offset_y_pt = (target_height_pt - image_height_px * scale) / 2
- x_pt = offset_x_pt + x_px * scale
- y_pt = offset_y_pt + y_px * scale
- w_pt = w_px * scale
- h_pt = h_px * scale
Use stretched conversion only for a ratio-matched canvas:
- x_pt = x_px * target_width_pt / image_width_px
- y_pt = y_px * target_height_pt / image_height_px
- w_pt = w_px * target_width_pt / image_width_px
- h_pt = h_px * target_height_pt / image_height_px
When exact image dimensions are known, use scripts/coordinate_helper.py to calculate scale values and examples.
Use named constants for canvas width, canvas height, and repeated colors.
Keep related objects grouped in code sections matching visual regions.

4. Code Generation

Generate a complete runnable VBA macro, not a fragment. For the default presentation target, also save the macro as a .bas or .vba file in the working directory so it can be run or pasted without copying from chat.

Use a two-pass generation strategy. The two-pass split exists because layout errors are cheapest to catch before any styling is applied; do not skip Pass 1 to save time.

Pass 1 — Skeleton: emit one macro that draws every manifest row as a flat gray rectangle (or thin gray line for connectors) at its mapped point coordinates, with the element id printed inside as a label. No fills, no gradients, no shadows, no fonts beyond a default 10pt. The skeleton's only job is to confirm bbox positions, sizes, and arrow endpoints match the source. If the skeleton is visibly wrong, fix the manifest first, not the styling.
Pass 2 — Styled reconstruction: only after the skeleton matches the source, regenerate the macro with real fills, strokes, fonts, gradients, shadows, transparency, and ZOrder. Preserved raster crops enter at this pass via Shapes.AddPicture.

You may emit both passes as two separate Sub procedures inside the same .bas file (for example Sub BuildSkeleton() and Sub BuildFinal()), so the user can run the skeleton first and the styled version second without regenerating code.

The macro must include:

A Sub ... End Sub wrapper.
Host-appropriate setup:
- PowerPoint: create a new presentation or a new blank slide by default, set slide size, and clear only generated shapes if working inside an existing deck.
- Excel: create or use a worksheet drawing area, clear shapes in the target area, optionally adjust cells behind it.
- Word: create or use a document page/canvas, clear generated shapes by name prefix where possible.
Canvas cleanup code.
Canvas size setup code.
Background creation code.
Code for every visible object, in back-to-front order.
Shapes.AddPicture code for each preserved local crop in hybrid mode, with LinkToFile:=msoFalse and SaveWithDocument:=msoTrue where the host supports it.
Named objects, for example shp.Name = "Header_Background".
Shared color helpers using RGB(...) or small helper functions.
Comments that label each visual region.
TextBox creation for legible image text.
Z-ordering when overlap matters, using ZOrder or creation order.
A stable generated-object prefix such as AITVBA_ for shape names and cleanup.

Use Office APIs conservatively:

Shapes.AddShape, Shapes.AddTextbox, Shapes.AddLine, Shapes.BuildFreeform, FreeformBuilder.
Shapes.AddPicture for preserved raster crops in hybrid mode.
Fill.ForeColor.RGB, Fill.Transparency, Line.ForeColor.RGB, Line.Weight.
TextFrame / TextFrame2 font settings as appropriate for the host.
Shadow, GradientStops, PresetGradient, or layered translucent shapes to approximate effects.

5. Presentation Materialization

For the default presentation target, after generating the VBA:

Save the generated code to a local .bas or .vba file.
If hybrid mode is used, save preserved crop assets locally. Use scripts/crop_preserved_elements.py when crop coordinates are known.
Run scripts/vba_lint.py against the generated VBA file.
Detect the available local presentation environment:
- Use scripts/detect_office_environment.py when possible.
Try to create/open the editable deck automatically when a supported path exists:
- On Windows with Microsoft PowerPoint installed, use scripts/run_powerpoint_vba_windows.ps1.
- On macOS with PowerPoint installed, use scripts/run_powerpoint_vba_macos.py when applicable.
- With WPS only, do not assume automatic macro injection. Open WPS when possible, provide the .bas path, and give WPS-specific manual run steps.
- If automation is blocked by Office macro security, AppleScript permissions, PowerPoint Trust Center, unsupported VBA injection, or WPS macro limitations, open the best available presentation app when possible and provide the exact macro file path plus concise manual steps.
Confirm in the final response which app was detected, which app was opened, whether the macro executed, which preserved assets were used, and whether a deck file was created. If it is not open, state the blocker directly.

The final answer must not imply that a PNG preview is the editable deliverable. It must distinguish between source image, preview screenshot, VBA source file, editable deck file, WPS-compatible file, and any non-VBA PPTX fallback.

6. Mandatory Self-Checks

Perform these self-check rounds before final output:

Structure check: verify no major object, region, text block, connector, or layer is omitted; verify back-to-front order.
Editability check: verify that text, arrows, boxes, connectors, and other structural elements were recreated as editable Shapes unless explicitly placed in a preserved crop.
Preservation check: if hybrid mode is used, verify that every preserved raster element has a crop asset, asset path, AddPicture placement, shape name, and reason for preservation.
Code check: verify VBA syntax, variable declarations, object references, Office host APIs, cleanup logic, shape names, generated-file path, and preserved asset paths. Use scripts/vba_lint.py as a lightweight smoke check when code is available as a file or can be copied into a temporary file; do not overstate it as a full VBA compiler.
Materialization check: verify that the macro was saved, that preserved assets were saved when needed, that the available office runtime was detected, that the appropriate PowerPoint/WPS path was attempted, and that the output deck is open or a concrete blocker is reported.
Visual check (render-verify 强制闭环): 不允许停留在"人工目测"。这是保证还原度的关键环节，必须按下面步骤反复迭代直到达标或触发硬性上限。

每一轮的固定动作：
1. 运行最新的 VBA 宏，生成 / 更新本地 deck。
2. 把当前幻灯片导出成 PNG（Windows 用 Slide.Export；macOS 用 AppleScript / PowerPoint 导出）。
3. 调用 scripts/compare_images.py 与原图比对，输出整体 SSIM、整体像素差比例，以及差异最大的若干局部区域（按 bbox 列出）。
4. 把每个高差异区域映射回 manifest 行，明确标注 "该区域对应 E03/R01/L05 的位置/颜色/尺寸/箭头终点偏差"，再针对性修改 manifest（不要漫无目的地全部重写）。
5. 重新生成 Pass 2 VBA，回到第 1 步。
通过条件（同时满足才能交付）：
- 整体 SSIM ≥ 0.90，且
- 没有任何高差异区域占画布面积 > 5%，且
- manifest 里所有 endpoints 字段对应的箭头在渲染图里目测连接到了正确的元素边界。
迭代上限：最多 6 轮。每轮必须在最终报告里附一行"第 N 轮：SSIM=0.xx，最大差异区域=…，本轮修改了 …"。

终止策略：
- 若 6 轮内达标，停止迭代，输出 deck 和报告。
- 若 6 轮仍未达标，但每轮 SSIM 持续上升（趋势明显），允许再扩 3 轮（共 9 轮上限),并在报告中说明仍在收敛。
- 若连续 2 轮 SSIM 不再上升或反而下降,立即停止,在报告里列出仍不达标的具体 manifest 行,而不是默默交付一个差的版本。
- 用户在对话中显式说"够了 / 先这样 / 不用再迭代"时，提前停止并说明当前最终 SSIM。
报告语言：每轮的差异分析、区域定位、修改说明都用简体中文（遵循文件顶部 Output Language 规则）。

仅在本机确实无法跑宏时（无 Office、Trust Center 拦截、沙箱环境），才允许跳过该闭环并降级为标注清楚的"人工目测估计"，且必须在报告中说明为什么没跑、用户在哪里能自行运行该闭环。

If a check finds a gap, revise the VBA before responding. Mention unresolved uncertainty only after revising what can be revised.

7. Output Requirements

Use the structure in references/output-template.md:

Short explanation of the reconstruction strategy.
Artifact paths for the VBA file and deck file/open state.
Preserved raster asset paths and a short editable-vs-preserved element summary when hybrid mode is used.
Complete VBA code block.
Brief steps for running the VBA.
Self-check results.
Possible visual errors and why they remain.
Next-round optimization suggestions, especially if the user can provide a rendered screenshot.

Always include a short "How to run" section. Keep it host-specific and practical.

8. Error and Fidelity Policy

Do not promise perfect pixel-level reconstruction for complex photos, textures, real people, dense raster effects, or highly complex gradients. Explain that Office Shapes can approximate these but cannot fully reproduce raster detail.

Unless the user explicitly allows full-image insertion, do not place the original image on the slide/sheet/document and claim it is reconstructed. This restriction does not prohibit small local preserved crops in hybrid mode, as long as they are clearly labeled. If using a full background image is appropriate, label it clearly as the background-assisted option and keep editable overlays separate.

Never replace the editable deliverable with a PNG, screenshot, or generated raster image. A raster image may be used only for source analysis, validation, optional preview, or explicitly labeled local preserved elements in hybrid mode.

For "1:1" requests:

Increase detail and object count.
Preserve aspect ratio, palette, and layer order.
Avoid inventing hidden or unreadable details.
Mark uncertain text or shapes with comments in the VBA.

Bundled Resources

scripts/vba_lint.py: lightweight checks for generated VBA.
scripts/detect_office_environment.py: detect likely Microsoft PowerPoint/WPS runtimes and recommend a materialization path.
scripts/crop_preserved_elements.py: crop local preserved raster elements from the source image for hybrid reconstruction.
scripts/run_powerpoint_vba_macos.py: attempt to run generated PowerPoint VBA through macOS PowerPoint automation and bring PowerPoint forward.
scripts/run_powerpoint_vba_windows.ps1: attempt to import and run generated VBA through Windows Microsoft PowerPoint COM automation.
scripts/compare_images.py: compare source image and rendered VBA screenshot.
scripts/coordinate_helper.py: convert image pixels to Office point coordinates.
references/vba-shape-patterns.md: common Office Shapes patterns.
references/reconstruction-checklist.md: detailed review checklist.
references/output-template.md: final response template.

About this skill

Automatic

Coding pack

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v5 — 9/3/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

xiao24bei / xiaobei-skill

skills/xiaobei-skill-image-to-vba/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 9/3/2026Version v1Bundled 12 files · scripts

Illustrative Want to see what xiao24bei/xiaobei-skill-image-to-vba would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Other skills for the same job

Published skills whose descriptions closely match this one, with their headline benchmark lift.

xiao24bei/xiaobei-skill-rebuild-image-in-powerpointUse when a user wants Codex to directly operate PowerPoint and reconstruct an entire reference image as editable PowerPoint objects while preserving its full layout, labels, connectors, and visual relationships. Do not use for paper-to-deck generation or VBA-only output.Close match+27 pts pass-rate lift

Related skills

Skills that do similar work, with their headline benchmark lift.

From the same source and nearby in the catalogue.

Ready to supercharge your AI agents?

SkillScore

Safety

What it changes

Track record

Use this skill

XiaoBei Image to VBA

Output Language

Purpose

Default Deliverable Contract

Hybrid Preservation Mode

保留元素的几何契约 (Preservation Geometry Contract)

保留元素自检 (额外的第 3.5 项 self-check)

Environment Compatibility

Workflow

1. Input Recognition

2. Image Parsing

2.5 Element Manifest (mandatory)

3. Coordinate Modeling

4. Code Generation

5. Presentation Materialization

6. Mandatory Self-Checks

7. Output Requirements

8. Error and Fidelity Policy

Bundled Resources

SkillSafety

Source

Other skills for the same job

Related skills

More in the registry