Ready to supercharge your AI agents?

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

gtrk-travel-recap（旅拍解说 · 组合成片图纸） · gitruck | DecimalAI

Design & Mediaby gitruck

gtrk-travel-recap（旅拍解说 · 组合成片图纸）

旅拍解说一站式成片图纸——输入一个旅拍素材文件夹（长视频或素材集），AI 理解素材、写三段式解说稿、一次确认后全自动跑完配音/建工程/拆分/B-roll/字卡/BGM/字幕，直达客户端可出片工程。当用户想「旅拍解说 / 旅行视频解说 / 把这条旅行长视频做成解说视频 / 旅拍素材做成片 / 长视频二创解说 / vlog 解说化 / travel recap」时使用本 skill。凡涉及把旅拍素材（自拍或网络长视频）做成中文解说成片，优先用本 skill 编排 gtrk CLI 全链，不要手搓单命令、别让用户逐步驱动。

Maintained

SkillScore

SkillScore82/100

Test results 82

Safety

3/3 safety checks passed

Code scan
Intent
Content

What it changes

14%82%of test cases passed +68 pts

Without skill3 of 22

With skill18 of 22

15 fixed3 passed either way4 still failing0 broke

Measured on gemini-3.8-flash11/6/2026v1

Track record

Agent runs: 0; All time

See one task, with and without the skill

One test task在对旅拍工程执行主轨 B-roll 自动铺设（gtrk matrix lay）时，为了在候选不足时优先放宽降级填满画面而非直接留黑片， CLI 命令行中必须显式指定哪一个填充策略参数？

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 gitruck-gtrk-travel-recap
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("gitruck-gtrk-travel-recap")
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 gitruck-gtrk-travel-recap

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/gitruck-gtrk-travel-recap/SKILL.md https://app.decimal.ai/s/gitruck-gtrk-travel-recap/SKILL.md

Copy SKILL.md

Save it in your project, as

.claude/skills/gitruck-gtrk-travel-recap/SKILL.md

This gets SKILL.md only. The skill's 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

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

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
from agents import Agent, Runner

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
from openai import OpenAI

client = OpenAI()
skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
import anthropic

client = anthropic.Anthropic()
skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/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/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
from langchain.agents import create_agent

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

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

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
from pydantic_ai import Agent

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
from deepagents import create_deep_agent

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from agent_framework import Agent

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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("gitruck-gtrk-travel-recap")
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/gitruck-gtrk-travel-recap/SKILL.md

Give it to your agent

from pathlib import Path
from strands import Agent

skill = Path(".agents/skills/gitruck-gtrk-travel-recap/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 1 other file is 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 gitruck-gtrk-travel-recap

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/gitruck-gtrk-travel-recap/SKILL.md https://app.decimal.ai/s/gitruck-gtrk-travel-recap/SKILL.md

Copy SKILL.md

Save it in your project, as

.agents/skills/gitruck-gtrk-travel-recap/SKILL.md

This gets SKILL.md only. The skill's 1 other file is 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/gitruck-gtrk-travel-recap/SKILL.md

Read the summary as JSON

curl -fsSL https://app.decimal.ai/s/gitruck-gtrk-travel-recap.json

This gets SKILL.md only. The skill's 1 other file is not included.

Other ways to add it

Other setups docs

You don't need a DecimalAI account.

Version: v1, tested 11/6/2026
Source: gitruck/cli ↗
License: MIT

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

gtrk-travel-recap（旅拍解说 · 组合成片图纸）

把一个旅拍素材文件夹变成一条中文解说成片工程：理解素材 → 三段式写稿 → 检查点①一次拍板 → 配音/建工程/拆分/B-roll/字卡/BGM/字幕全自动 → 客户端出片。

定位：本 skill 是第一张 structure 级组合图纸——它不新增任何命令，只编排既有原子零件（transcript / matrix / project init / split / mg / audio lay / subtitle lay / TTS 工具）。零件的参数细节以各自 skill（gtrk-matrix / gtrk-splitter / gtrk-mg / gtrk-transcript）与 --help 为准，本图纸引用不复制；但编排次序、检查点、配方级参数以本图纸为准。CLI 是手，你是脑：写稿、切 beat、挑候选信号、产字卡颗粒、管检查点，都是你的活。

家族归位：本图纸是解说链（gtrk-narration 正本）的旅拍垂类实例。时序铺排、引用段、出处对齐等链级条款以 gtrk-narration 为正本；本图纸只补充旅拍要素层（贵大奇多钩子、攻略 tips、升华笔法）。素材无自带时序的题材（科普/情感/带货）走 gtrk-voiceover。

本 skill 已含跑通全链需要的编排信息。

一、输入契约与双场景识别

用户给一个素材文件夹（或多个，逗号分隔传 --dirs）：

内容	必须性	说明
视频/图片素材	必须	一条长视频（二创）或一堆素材（原创拍摄）皆可；图片会被自动运镜成动态上轨
BGM 音频文件	可选	没有就走素材矩阵搜索（见检查点①）
介绍类文本（txt/md）	可选	有则作写稿补充素材

一夹多片时检索域按场景分流：二创场景一个夹里常躺着好几条互不相干的片， B-roll 检索域 MUST 收窄到正在做的那一部片本身——--dirs 既吃文件夹也吃单个素材文件，传文件即把域收窄到该素材（见 §五 ⑦ 的★铁则）。原创（自拍素材集）那一档整夹就是这一片的素材域，传夹正确。

MUST NOT 下载二创源片 / 主体素材：网络视频二创场景下，用户提供的文件夹就是输入，下载动作归用户。平台授权素材（素材矩阵的 BGM / 空镜）可按其签名直链落地；这条例外不适用于源片与主体画面素材。

双场景识别（决定理解编排走法）：

二创：文件夹里是网络旅行长视频（通常含博主解说语音）→ 原片文案是写稿主料；
原创：用户自己拍的素材集（通常无解说语音）→ 素材理解（describe）是写稿主料。
拿不准就 ffprobe 看时长与条数：单条 10 分钟以上大概率二创；几十条短素材大概率原创。仍拿不准，问用户一句。
原创场景的素材集常常通篇没有对白——那就不跑转写，下表的语种一问以「无对白 / 不转写」闭合，MUST NOT 因缺这个答案阻塞开工。

开工三问：稿源 · 声音基座 · 源片语种。三问开工时与双场景识别一并问清、一张表一次问完、MUST NOT 代猜：

问	选项与走法
稿从哪来	AI 代写（缺省，§三规约写）/ 用户自备稿——跳过写稿，稿直接进检查点①过目（文风铁律不强加于用户手稿，只提示明显硬伤）
声音怎么来	TTS 配音（缺省，检查点①挑音色）/ 用户自己念——拿念稿毛片先走 `gtrk oralcut` 粗剪（带稿最准；外录收音先 `gtrk audio align` 对轨），产出的粗剪工程即时间基座，§五的「配音④→建工程⑤」两步跳过，从拆分⑥接入，后续铺排/字卡/BGM/字幕照旧
源片语种 / 方言	取证=听一段 / 看片内有无内嵌字幕 / 看素材出处；答案定 §五 ① 的 `gtrk transcript --lang`（`zh-CN` 普通话 / `zh-HK` 粤语 / `en-US` / `ja-JP`…）。MUST 问用户、MUST NOT 代猜、MUST NOT 以缺省 `zh-CN` 静默开跑转写——转写是计费动作且早于检查点①，非中文素材传错语种会让检查点①拿到错误稿件并产生重复 ASR 成本。粤语建议如实填 `zh-HK`；当前 `--json` 路径按字级转写。素材无对白：答「无对白 / 不转写」即闭合

推荐面（公约 §三‴）：问答收口时可问一句「要不要看看第三方 skill？」；题材命中场景触发词才定向推荐，读取 gtrk skills recommend --scene <id> --json；用户拒绝后不执行推荐动作。

精剪包装（公约 §三⁵）：开工时告知一句「快速成片会在适合的段落做合成包装，最多全片四分之一，不要就说一声」——不是问题，用户说不要即零动作；gtrk composite bake --abi 报烘焙浏览器没装时，在这张问询表里多问一句装不装（不装用本机 Chrome / Edge）。

一′、素材适配闸（写稿前必过）

本闸判的是题材，与「有无自带时序」正交。剧情短片、摆拍表演和单一室内场景不能直接套用旅拍要素层。

取证零新增计费（读的都是编排里本来就要产的东西，MUST NOT 为这道闸新加一次 describe 或转写）：

原创场景：读 gtrk matrix index / gtrk matrix describe 已产的 desc（中文视觉描述）与 tags；
二创场景：读 gtrk transcript 全文。

判据（可判定，逐条素材过一遍）：

主判据——这一条能不能回答「这是在哪、看到了什么」（可辨认的地点 + 所见之物）。能答的占比过低 ⇒ 不适配；
反向命中任一即不适配：文本呈现剧情对白（人物一来一往推进情节）/ 摆拍表演（演出来的而非记录下来的）/ 单一室内场景（全片不出一个可辨认的地点）。

命中即硬停：MUST NOT 进入写稿（形制照 gtrk-vlog-docu 的立神闸）。硬停位置是写稿之前 = TTS 计费之前——它与 §四检查点①「未经确认 MUST NOT 发起计费动作」是同一条防线的前置一层：检查点①拦的是「稿子对不对」，本闸拦的是「这活该不该接」。停下时 MUST 如实列出哪几条不适配 + 依据原文 （引 desc / tags / 转写里的原句，别只下结论），并给三条出路：

出路	怎么走
① 剔除	把不适配的那几条摘出去，用剩下的素材继续跑本图纸
② 改投	剧情 / 影视 / 游戏这类有时序但非旅拍的题材走 `gtrk-narration`（解说链正本）；素材无自带时序的题材（科普/情感/带货）走 `gtrk-voiceover`
③ 照跑	用户坚持就照办，但 MUST 在交付说明里如实登记「素材与旅拍垂类不匹配、要素层（贵大奇多钩子 / 攻略 tips）多半落空」（降档路径如实登记，口径同《成片型图纸公约》）

两条边界写死：agent MUST NOT 自行判死（决定权在用户，本闸只提名不宣判）、MUST NOT 静默照跑—— 静默照跑会让整条链在错误题材上完成并产生不必要的计费，因此必须在此停下。

二、全链总览（快速成片）

① 理解        transcript 原片（二创）/ matrix index + describe（原创）
①′ 适配闸【必过】读 ① 的既有产物判题材（见 §一′）——不是旅拍就硬停，MUST NOT 进入写稿
② 写稿        三段式规约（见 §三）→ 解说稿 + 标题 + HashTag
③ 检查点①     【必停】稿件确认 · 画幅 · 音色 · BGM（gtrk matrix material --scope audio）· 字幕样式 · B-roll 编排量预估（§四）
④ 配音        gtrk tool audio_tts_clone（≤5000 字，句级时码直出）
⑤ 建工程      gtrk project init --tts-task <id> --canvas <WxH>
⑥ 拆分派单    gtrk split 投影 → 你写拆分稿（含字卡 overlay aux）→ split 落地
⑦ B-roll      matrix --local --dirs（二创钉单片）→ describe --plan → lay --mark-weight 0.3 --yes
⑧ 抽帧检查    对字卡挂点抽帧看底轨构图（gtrk-mg SOP）
⑨ MG 字卡     按 §六 轻量规约产颗粒 → mg lint → gtrk mg --project
⑩ BGM        gtrk audio lay --volume 0.1 --beat-align（-20dB 口径）
⑪ 字幕        gtrk subtitle lay --style <检查点①拍板样式>（用户没选用 default）
⑪′ 精剪包装   交棒 /gtrk-composite（公约 §三⁵：auto ≤ 25% → 复核 → 构建机检 → 铺轨烘焙，不停）
⑫ 交付        客户端打开工程出片（gtrk render 仅主轨快照预览，见 §八）

检查点①之后一路跑完不再打扰用户——垂类示例的承诺面就是快速成片一杆到底。 ⚠️ 这条承诺以 ①′ 适配闸通过为前提：闸命中时的那次停顿不算「打扰用户」，它是在把钱花出去之前问一句「这活该不该接」。交付后对个别环节不满意走 §七增量重跑，对垂类玩法本身不满意走 §七个性化出路。

三、三段式写稿规约

通用框架层（可移植——其他垂类图纸原样复用本层）

结构三段式：第一部分「钩子」抓眼球保留存 → 第二部分「干货」给内容价值 → 第三部分「升华」给情绪价值。爆款四属性贯穿：快乐+知识+共鸣+节奏。

背景检索建议：写稿信息源除了画面与音频（字幕不一定有），建议执行 agent 对素材主题的关键词做一轮背景/上下文检索（地名典故、店家来历、片名与公开报道等），检索所得须与素材内可见事实互证后才可入稿，并标注来源。有检索能力就做、没有不强求（能力缺席按纯素材信息写，不算违约）；检索到的外部数字与争议性说法宁缺勿滥。

文风铁律：

禁模板连接词：让我们/首先/然后/最后/此外等一律不用；
拒翻译腔：全部短句中文语感，无欧化从句；
禁「否定—转折」对举句式（最高优先级）：禁的是整个家族，不只字面：
- 「这不是 X，而是 Y」／「与其说是 X，不如说是 Y」
- 「X 不仅仅是 Y，更是 Z」／「X 不只是 Y，它还是 Z」
- 「重要的不是 X，而是 Y」／「真正的 X 从来不是 Y」
- 「不是因为 X，而是因为 Y」／「X 靠的不是 Y，是 Z」
- 识别特征：先立一个没人主张过的靶子再推翻它、用对偶结构制造顿悟感、句子的信息量全在后半句而前半句只是垫脚。
- 改法：把后半句直接说出来，删掉前面那个假靶子。「这不是一条水渠，而是一条生命线」→「几百个村民用铁锤在悬崖上凿了三年，就为这一条水渠」。能直接陈述的事实，绝不用对举包装。
- 这条也适用于标题与 HashTag。
禁幼稚拟声与过度儿化（★ 打样校准）：「轰隆隆砸下来」这类拟声，换成真实事实（「从九十多米高处坠进谷底，比尼亚加拉还高出近一倍」）——数字与事实比拟声词有力得多；
风格目标 = 与原素材旗鼓相当：原片多吸引人，解说就得配得上——贴着原片气质走，不加分（加分是业务化的事，走栏目风格）不减分。

成片时长 = min(内容所支撑, 素材所支撑)：

内容支撑：有多少真东西说多少话，素材再多也不注水；
素材支撑：按「30% 图片×8 秒 + 70% 视频素材时长」估算上限，稿子在写稿阶段就按上限收敛（不要等铺轨时开天窗）；把这个秒数换算成字数用下一条的折算口径（公式本身不动）；
折算口径 = 挂钟语速（wall-clock：稿件字数 ÷ 合成音频总时长，含句间停顿），按检查点①选定的那条音色取值：读 catalog 该音色的 speech_rate_cpm（字/分钟），并把 speech_rate_basis 一并报给用户（measured = 实测 / derived = 由声学参数推算）；catalog 没给该字段（非中文音色）或拉不到 catalog 时才用缺省 ≈330 字/分钟，并明说这是缺省估值、不是该音色的实测值。3 分钟成片 ≈ 1000 字。缺省 330 的依据与标定方法见 references/speech-rate.md，本图纸只引用该单一真值源。
⚠️ MUST NOT 拿计费折算常量（240 字/分钟）推稿子字数：计费口径与写稿口径分开，计费常量只用于估算费用。

写后自校验清单（逐项过，不过重写）：

三段结构完整、时长在 min 公式内
禁词零命中、无翻译腔、无幼稚拟声
「否定—转折」对举句式零命中（逐句扫「不是…而是」「不仅…更是」「与其…不如」及其变体；含标题与 HashTag）
重复字/错字扫描（★ 打样实锤：「下瀑布瀑布」重复字 TTS 会照念）
事实矫正：转写稿的明显错误（如数量级离谱的数字）按常识矫正或模糊处理，不确定的不引用
解说序 ≈ 素材序（二创场景：按原片游览顺序写，B-roll 时间窗天然可用）
同批产出：标题 3-5 个（≤30 字）+ HashTag 3-5 个

旅拍要素层（垂类专属）

钩子要素（第一部分重点找）：贵大奇多 / 热门 IP / 名人名著 / 知名历史事件 / 全球 TOP 榜 / 奇闻轶事 / 脍炙人口；
干货要素（第二部分）：攻略 tips / 游览流程 / 干货&经验 / 旅途见闻，幽默口吻、隔几句一个梗（谐音/反转/无厘头），在精不在多；
升华笔法（第三部分）：思考/感悟/启发/感动，正能量收束；第一人称历数、物象隐喻收尾都是成熟笔法；
有文化衍生空间（历史典故/风俗人文）可适当发挥，但以原素材内容为锚。

三′、L1 看点 rubric · 画面投影版（供 `--highlight-weight` 使用）

highlight 是镜头看点分——服务端 VLM 对着一帧画面打 0–100。旅拍这一档的准则（highlight_rubric）：

判这一帧对旅拍短视频有没有「看点」，0-100：
- 奇观地貌：雅丹/冰川/星空/沙漠/峡谷等罕见地形
- 极端天候光线：日出日落、逆光、雪雾、暴雨、极光
- 人文场面密度：市集、节庆、宗教仪式、当地人劳作
- 反差罕见：人与巨大景观的比例、无人区里的人造物、动物
- 规模感：能看出「有多大/有多高/有多远」的构图
低分：车内挡风玻璃前方公路、酒店房间、无主体的天空、拍糊的中间帧

分工注记（MUST 遵守）：本 rubric 只判镜头可见之物。知识侧要素——地名典故、历史事件、人文背景——不进画面 rubric，它们归写稿与背景检索（§三通用框架层已列）。

分工是：画面 rubric 管「这一帧值不值得看」，写稿管「这件事值不值得讲」。 平淡的过渡帧判低分不等于不可用——降权不排除，候选稀疏时照样上（留空更难看）。

传入方式：把上面代码块原样存成工程目录下的 highlight-rubric.txt，理解那一步带上它——

gtrk matrix describe --plan <工程>/split/broll-plan.json --highlight-rubric @<工程>/highlight-rubric.txt

存成文件而非内联，除了免去多行转义，还留下「这条片用的是哪套准则」的可查痕迹。跑完 gtrk matrix lay 自动读 plan 里钉存的同一个桶，不必再传；真传了而且与 plan 钉的不同源，CLI 会硬失败（不静默择一）。 ⚠️ 换准则重新打分要重新看片（1 积分/张）：客观描述缓存不被覆盖，但看点分按准则分桶，新桶没分就得重打。

四、检查点①（双模式共有，必停）

写完稿必须停下，把以下六项一次给用户拍板。未经确认 MUST NOT 发起 TTS 与 B-roll 云端编排计费动作。

本地素材的编排缺省走云端并按编排量计费。matrix lay 跑前必须报告预估用量，取得同意后才能带 --yes；无 TTY 且未确认时整轮中止，工程不改动，plan 可重跑。

语种不在这六项里：它在 §一开工三问就已定，检查点①对它最多做过目复核。 MUST NOT 把语种拖到这一步才问——此时 ① 的 gtrk transcript 已经跑完并已计费，稿子也已按错语种写出来了。

解说稿 + 标题：正文全文贴给用户（不是给路径），标题候选列出；
画幅：问目标平台——横屏 B 站/西瓜=1920x1080，竖屏抖音=1080x1920。素材几何与画幅不一致时明确提示（横素材→竖画布会裁切或留边）；
TTS 音色：拉 catalog（GET {apiBase}/task/tts/voices，免鉴权）按内容调性挑 2-4 个候选，每个附试听链接（catalog 的 audition_url 字段，api.ai-mcn.tv:9000 静态直链）——用户在对话框里直接点开听。internal 标注的音色注明「仅内部成员」。也可按用户描述语义挑（「要个温柔女声」）。 每个候选还 SHALL 附上该音色的挂钟语速与据此推的稿件字数区间（取同一份 catalog 载荷里的 speech_rate_cpm + speech_rate_basis，零额外往返；该字段缺席就明说「用缺省 330 估」；标定方法见 references/speech-rate.md）——挑音色的同时也是在挑稿子长度，别让用户回头才发现稿子长了三成。这条是加项，MUST NOT 削弱既有的「无试听链接 = 未完成推荐」义务；
BGM 处置（双来源）：
- 用户有本地音乐 → 询问是否 gtrk tool audio_separation 伴奏分离后取伴奏（缺省建议分离——TTS 配音不与歌曲人声打架）；
- 用户未提供 → 搜同和素材矩阵：gtrk matrix material "<按稿件情绪+地理文化写的检索词>" --scope audio --top-k 5 --diversity --json （检索词口径见公约 §三′：内容驱动、禁通用词、至少含一个只有这一片才成立的词；候选 MUST 给用户挑不许默认 top1；跨片避让由 audio lay 记账 + 检索缺省 --exclude-recent 自动完成（按成片时长可加 --min-duration <秒>）。身份路由与计费口径由零件持有；版权状态要如实转述——is_copyright=true = 可商用、false = 不可商用。矩阵成员档缺省搜全库并逐条给 copyright_label；用户特意说明只要可商用或商用发布时才加 --commercial-only 收紧。upsell 是独立顶层字段，不是候选。推荐 3-5 首、每首附试听链接（结果的 download_url）；audio_type:"pure" 纯音乐直接用，"song" 直接取 accompaniment_url 现成伴奏（零处理成本，两档都有）；注意曲长 vs 成片时长（duration 字段）：audio lay 缺省循环叠满至工程末尾、尾段裁齐（多 clip 首尾相接，命令会打 INFO 报段数），短曲只是接缝多一些、不会铺一半就静音（想少几道接缝，挑曲时加 --min-duration <秒>）；要单次铺请显式传 --no-loop；
- 试听链接义务（MUST）：音色与 BGM 的推荐没有试听链接=未完成推荐。
- BGM 落地一步（MUST）：检索出参只给签名 URL，而 gtrk audio lay --file 只收本地路径。拍板后先把选中曲落成本地文件：audio_type:"song" 取 accompaniment_url 伴奏版，"pure" 取 download_url。文件名按 bgm-<曲名>[-伴奏].<扩展名>，供跨片避让记账使用。
字幕样式：7 种样式（default/outline/cinema_yellow/immersive_box/wide_spacing/deep_shadow/boxed）× 11 色（雅黑/淡绿/森林绿/湖蓝/道奇蓝/钢蓝/浅粉红/深橙/珊瑚橙/橙红/土豪金）挑选；用户没选就用 default+雅黑（不追问不阻塞）。拍板值在 ⑪ gtrk subtitle lay 落地，客户端里仍可整轨换样式（悔棋通道）。
B-roll 编排量预估（本地素材路必报）：本地素材产的 plan 走 matrix lay 时缺省即云端编排并按编排量计费（--arrange local 不受理）。先拿数再拍板：
```
gtrk matrix lay --project "<工程目录>" --arrange-estimate-only --json   # 停在计价确认之前：零云端调用、工程文件零改动
```
把它与 TTS / describe 的预估同屏报给用户，一次拿到同意——§五 ⑦ 的 lay 才允许带 --yes。报不出这笔就不许开跑。

五、快速成片编排（打样验证过的真实命令）

产物落点纪律（MUST · 全文见随包 AGENT.md 同名一节）：文本产物写入 CLI 返回路径或用户显式指定的 -o, --out；交付物直接写入工程目录；引用媒体用原路径，不复制到 agent 工作目录；临时文件放系统 temp，用完即删。交付物 MUST NOT 先写 agent 自有目录再拷进工程；中转本身即违规，体积不是豁免理由。

# ① 理解（二创：转写原片，--json 顺产 transcript.json 备用）
gtrk transcript "<原片>" --lang <zh-CN 普通话 | zh-HK 粤语 | en-US | ja-JP …> --json -o "<工程目录>/transcript/<片名>-transcript.md"
#   ↑ 产物落点：不带 -o 时两个产物都落**源片同目录**（`<名>-transcript.md`；--json 另产 `<名>-transcript.json`，
#     .json 跟随 .md）。缺省本身是在案设计，但二创素材多在用户素材库里 ⇒ SHOULD 用 -o 挪进工程目录（见本节顶部纪律块）
#   ↑ 语种取开工三问的答案，MUST NOT 拿缺省 zh-CN 顶上去 —— **但这条对「非中文」素材才有钱可省**：
#     粤语 SHOULD 如实填 zh-HK（元数据诚实 + 备将来链路切换），当前链路下它与 zh-CN **对粤语无差别** ——
#     CLI 写死 word_level: true（src/commands/transcript.ts:178）⇒ 链路钉在豆包、到不了 Whisper
#     （infra utils/process/media/audio/asr.py:44-49 在 AsrFunc.Whisper 分支里仍把 word_level_handler 钉成豆包）；
#     而豆包字典里 zh-CN 与 zh-HK **都映射空串**（utils/partner/volcengine/doubao_llm_asr.py:36-37）
#     ⇒ 两种写法发出的请求逐字段相同。粤语能识别对，靠的是豆包空串模式自己的方言自适应，与这个参数无关。
#     "zh-HK": "yue" 那条映射真实存在（utils/partner/openai/__init__.py:18），但只在 word_level=false
#     走 Whisper 时生效，CLI 今天走不到那条路。
#   ↑ **MUST NOT 以「粤语语种传错」为由重跑转写**：la 进 ASR 缓存键（asr_result_cache.py:71-75），
#     换码必然 cache miss，白扣一次时长换回逐字段相同的请求与同一份结果。
# 原创场景改跑：gtrk matrix index --dirs "<素材夹>" && gtrk matrix describe --materials ...
#   ↑ **这个 `&&` 现在真的会拦住**：全域零枚举（一个素材都没枚举到）是**硬失败**（退出码 1），
#     `describe` 不会执行 —— 这是改对了：以前 index 报 0/0 也 exit 0，`describe` 会对着空清单空跑一趟。
#     报 0/0 先看两件事：路径里有没有**英文半角逗号**（有就把 --dirs 重复传 `--dirs "A" --dirs "B"`，累加不覆盖）、
#     素材夹里有没有断链。⚠️ 「个别素材处理失败」不算硬失败（那是既有 failed 计数口径，退出码仍 0）。
#   ↑ 原创（自拍素材集）这一档**整夹即该片素材域，传夹正确**，不受 ⑦ 那条「钉单片」铁则约束——
#     那条治的是二创：一个夹里躺着几条互不相干的片，才会互相抢候选

# ②③ 写稿（你执笔，§三规约）→ 检查点①拍板
#   BGM 候选（用户没自带音乐时，检查点①之前搜好一起拍板；试听链接=结果的 download_url）
gtrk matrix material "<情绪+地理文化检索词>" --scope audio --top-k 5 --diversity --json

# ④ 配音（音色 code 用 catalog 的 voice_id；稿子存 txt 传 --text-file）
gtrk tool audio_tts_clone --text-file <稿.txt> --speaker <voice_id> --json

# ④′ 语速回校（零成本自我标定；快速成片照跑不停顿，只汇报）
#   实测挂钟语速 = 稿件字数 ÷ 音频实际时长（时长取 ffprobe duration；字数口径同服务端：Unicode 码点、含标点与英文数字）
#   与写稿时所用的估值比对：**偏差 >15% MUST 在交付汇报里点名**（报「估 X / 实测 Y / 偏 Z%」），
#   并按 references/speech-rate.md 的五项格式追加本片标定记录，供后续估值复核。
#   MUST NOT 自动改稿、MUST NOT 自动重跑 TTS：改稿得从 ④ 重跑（TTS 变=全链时码变，⑤起全部重建）、要重付 TTS 费用，返工与否是用户的决定。

# ⑤ 建工程（画幅按检查点①拍板；产物目录含 gtrk/transcript/audio）
gtrk project init --tts-task <TTS任务id> --canvas 1920x1080 --no-open --json
#   ↑ 产的 transcript 文稿缺省去逗号句号（成片字幕口径）；要**带标点的文稿版本**
#     （回头改稿、转投他处、给人读）加 --keep-punctuation，工程结构与时码不受影响

# ⑥ 拆分：先投影拿句表，再写拆分稿，再落地
gtrk split --project "<工程目录>" --json          # 投影：拿 utterance id/时码/hash
# 你写 split/visual-split.json：beat 全覆盖（配音工程无 A-roll 画面，留隙=黑屏！）、
#   lane 全 FILM_BROLL、base_track="旁白主导"、narrative 用八枚举、
#   字卡挂 aux_layers type:"overlay"（handoff.category:"overlay"+duration_hint，进 dispatch.mg）、
#   看点关键词圈 handoff.anchors（[{keyword,utterance,query}] ≤2 个/beat——地名/奇观/数字头衔类实词，
#   铺轨把锚 query **原始 sim 最高的合格命中**（原始检索相似度，不是 mark/highlight 融合分）
#   钉在关键词说出时刻：听到「野牛高速」的瞬间看到野牛；锚预留优先于普通序贯槽消费，
#   取不到第一名时锚 outcome 如实报名次与去向；圈定指引见 gtrk-splitter）
#   ⚠️ 上面这串是**非穷举**摘录（另含 irreplaceability / rhythm / visual_task / container_stage 四个必填项）：
#   beat 必填字段全表以 gtrk-splitter 的 references/field-schema.md 为准，MUST 照该表写全
gtrk split "<拆分稿>" --project "<工程目录>" --json
#   ↑ 落地后走一遍 §六 末的**字卡自检清单**：数 type:"overlay" 的 aux 条数、核对 --json 回执里
#     queues.mg 的条数是否落在「每片 2-4 颗」

# ⑦ B-roll 三连（describe 给字卡抽帧检查与 mark 加权供料；internal 用户 describe 免费）
# ★ 铁则（口径同 gtrk-narration 与 gtrk-matrix，不许长出第二种说法）：解说链**一稿对一片**——
#   二创场景 --dirs MUST 钉到**这一部片的绝对路径**，**不许传素材夹**。传文件就是把检索域收窄到该素材；同夹多片时一次实测 187 条候选中有 86 条来自邻片（46.0%），不能把邻片结果当作次优解。
gtrk matrix index --dirs "<这一部片的绝对路径>" --json
#   ↑ 二创也要先建索引（原来 index 只挂在原创分支下，于是只能顺手对整夹建、正好把三条片灌进同一个 index.db）；
#     如果整夹已有索引，**不必重建**，检索侧传单片即可收窄（判据是路径前缀，传的就是那一条素材本身）
gtrk matrix --project "<工程目录>" --local --dirs "<这一部片的绝对路径>" --lay 0 --json
gtrk matrix describe --plan "<工程目录>/split/broll-plan.json" --yes --json
#   ↑ describe 结果里的 usable_flags 由你裁定（text_overlay 命中=撞原片字幕区,考虑换段或保留交客户端悔棋）
#   ↑ 同时读 describe 输出的**剔除风险**清单：候选数 ≤1 的 beat 一删就掏空
#     （音频驱动 ⇒ 整段黑屏或跨 beat 借片；口播 ⇒ 露 A-roll）。先补素材再剔，MUST NOT 见 flag 就删。
#     候选稀疏时先补素材再剔除，按 flag 直接删除可能掏空整段
gtrk matrix lay --project "<工程目录>" --mark-weight 0.3 --highlight-weight 0.2 --gap-fill fast --yes --json
#   ↑ --yes 的前提是**编排量预估已在检查点①（§四.6）报过并拿到同意**；MUST NOT 未经同意就带 `--yes`。
#     不带 --yes 且无 TTY 时本步**整轮铺轨中止**、工程零改动（plan 照常可用、可重跑）——
#     未报预估或未取得同意时不要继续执行

# ⑧⑨ 字卡（§六 规约产颗粒 → lint → 铺）
gtrk mg lint "<工程目录>/mg/<composition_id>.html" --dispatch "<工程目录>/split/dispatch.json"
gtrk mg --project "<工程目录>" --json

# ⑩ BGM（音量口径见下；<bgm> 是**本地文件**，检查点①拍板后先落地——URL 传给 --file 必炸）
# BGM 是带人声歌曲（audio_type:"song"）→ MUST 下载 accompaniment_url 伴奏版上轨，
# 不得用 download_url 原曲（人声与配音会打架）；"pure" 取 download_url。
# 文件名 bgm-<曲名>[-伴奏].<扩展名>：跨片避让按**标题归一化键**记账（剥 bgm- 前缀与伴奏/instrumental/off-vocal 后缀），乱命名=下一轮 --exclude-recent 失准
# --beat-align 的高潮点取自 split.beats 的 narrative / container_stage（升华段 → 容器转折 → 回扣段）；
# 拆分稿没带这两个字段 ⇒ 落 0.75×全片兜底档，**CLI 会明示那是猜的**。不接受就用 --climax <轨秒> 指定。
# 两侧都够长时零平铺（恰好 1 个 clip）；不够长才按小节线平铺补齐、接缝吸附 downbeat。
gtrk audio lay --project "<工程目录>" --file "<bgm>" --volume 0.1 --beat-align --json

# ⑪ 字幕（样式/颜色 = 检查点①拍板值；用户没选用 default/雅黑；零计费——拆行走云端 subtitle_line_split 0 积分留痕，离线可加 --offline 用本地拆窗、重跑幂等替换）
gtrk subtitle lay --project "<工程目录>" --style default --color 雅黑 --json

# ⑪′ 精剪包装：见下方引注

⑪′ 精剪包装：交棒 /gtrk-composite 做精剪包装（公约 §三⁵：快速成片 auto ≤ 25% → 逐窗口复核 → 构建机检 → 铺轨烘焙，不停；回执列「做了 / 建议了没做 / 烘焙结果」；逐步推进按合成 skill 两个检查点停；开工时用户说了不要就跳过）。旅拍主画面多是素材本身，按「没有讲者」口径只用 B-roll 与排版。

停顿太长时：自训音色加 --fragment-interval 0.2（合成时就对）；云引擎音色不支持该参数（传了会报错，不会静默忽略），改在合成后跑 gtrk audio tighten --project <工程> 收紧句间停顿（纯本地零计费；只压跨句界的停顿，句内换气与原声引用段不动）。

配方级参数口径：

--mark-weight 0.3：美观度加权起步值（mark 缺失素材中性不受罚）；
--highlight-weight 0.2：看点加权，配 §三′ 的 rubric—— 须在 describe 那一步带 --highlight-rubric @<工程>/highlight-rubric.txt，不带就是按服务端缺省准则打分，旅拍这一档的奇观地貌/极端天候/规模感全不算数。与 mark 正交—— mark 判「好不好看」、highlight 判「有没有看点」。两权之和 0.5，语义仍占一半， MUST NOT 加到语义压不住（那会变成「专挑好看的，不管说的是什么」）。 ⚠️ 两个权重与 blurry 降权都依赖 matrix describe --plan 跑过（编排 ⑦ 那一步）；没跑就是全部候选中性、排序与不开权重逐字节一致——lay 会就此打告警，看到就说明漏了；
--gap-fill fast：主轨空洞按五级依次填——放宽地板取剩余候选 → 段界内延长相邻颗粒 → 跨 beat 借未消费候选 → 次地板补真画面 → 仍不够才落黑片。借来的画面与本段稿子相关性可能较弱，次地板槽可能很短；两者都会在日志里指名报出。CLI 缺省是 solid，本配方须显式传 fast；
句界吸附（--cut-align）默认 0.7，约七成字幕句起点贴近镜头切点，其余有意错开；
BGM 音量口径 = -20dB（线性 0.10）：BGM 作为垫底，volume 字段只写线性增益；客户端需要更响时再调整。
时间窗（--source-window）：二创场景解说序≈素材序时按 beat 对应的原片区间约束检索（可选增强，全库检索在几百场景规模下命中率已够）；
素材不二用、候选片段自带纯黑底与空档补黑、层序（common<concept<local，契约=大号在上；旧工程的黑底垫轨在最低号）均由 lay 内建，MUST NOT 绕过。

六、MG 轻量字卡规约（旅拍向最小集，不建栏目 skill）

本节采用「因势象形、不建栏目 skill」原则：风格锚定本片内容，不强行套用预设模板。原则正本见《成片型图纸公约》§三。

快速成片禁留白——字卡按下述规约直接产好叠上，用户不满意在客户端删，别等确认。

内容：地名字卡为主（地标名中文+英文/头衔副标），可加里程/价格数字卡；每片 2-4 颗、挂在地标出场句（拆分稿 aux_layers overlay）；
形态：左下安全区地名条——竖色条装饰+半透明深色衬底圆角条+主标题（60px 级粗体白字）+副标题（28px 级 75% 白）；透明叠加不挡画面主体；
动效：入场滑入淡入约 1.5-2s → 定格驻留到坑位末尾（gsap-emit v1 铁律⑦：时间线总长≥坑位+0.3s，禁全局渐隐）；
工艺：严格按 contracts/gsap-emit-v1.md 颗粒骨架（<template> 包裹/1920×1080/gsap.timeline({paused:true})/__timelines 注册/无随机无时钟/lib.baomitu.com GSAP CDN）；字体用渲染字体注册表在册族（如 Alibaba PuHuiTi）；实心衬底写子层不写根（铁律 4）；
产前抽帧（overlay 必做）：对挂点用本地 ffmpeg 抽 1 帧看底轨明暗与主体位置（临时文件落系统 temp 用完即删）；
产物存 <工程目录>/mg/<composition_id>.html，逐颗 lint 过了再铺。

字卡自检清单（形制同 §三「写后自校验清单」）：

拆分稿写完，数一遍 aux_layers 里 type:"overlay" 的条数，是否落在每片 2-4 颗；
落地后核对 gtrk split --json 的 queues.mg 条数，与上一条对得上、且仍在 2-4；
不足按地标出场句补挂；超出按「稀缺才有力」裁掉。

这是图纸层的审美自检，不是 CLI 通用硬闸：split 与 mg 负责报告实际条数，是否在 2–4 颗由本垂类图纸裁定。补在缺口上最省，零代码变更。⚠️ 反方也如实记：颗数是 dispatch.mg.length 一行就能算的机器可判量，若这类漂移反复发生，改上「通用可选入参 + 由图纸传值 + 非致命告警」是正当的。

七、增量调整与个性化出路（垂类示例定位）

本图纸是解说链的垂类示例，承诺面只有快速成片（检查点①一次拍板后一杆到底；决策前置三条腿照公约执行，单模式不豁免）。交付后对个别环节不满意，用下表增量重跑悔棋——改哪步跑哪步，不整链重来：

环节	暂停点产物	调整方式	增量重跑
写稿	解说稿 vN	用户改意见 → 你改稿	改稿后从 ④ 重跑（TTS 变=全链时码变，⑤起全部重建）
拆分	split/view.json + 拆分稿	改 beat 划分/queries	`gtrk split <拆分稿> --project` 重落
B-roll	broll-plan.json	编辑 plan（删候选/换段/pinned 钉选）	`gtrk matrix lay`（消费 plan 零检索；⚠️ 本地素材路的编排仍按编排量计费，不是零计费——重跑前照 §四.6 再报一次预估）
字卡	mg/*.html	改颗粒 HTML	`gtrk mg --project --only <beatId>` 单颗重铺
BGM	audio_track	换歌/换音量	`gtrk audio lay` 同文件幂等替换
字幕	cve 字幕 lane	换样式/颜色	`gtrk subtitle lay --style/--color` 重跑幂等替换（手加 text 不受影响）；逐条微调在客户端
精剪包装	`composite/plan.md` + 合成颗粒	不要某颗 / 补一段	客户端关合成轨可见或删那一颗；补做走 `/gtrk-composite` 指定区间（`plan --range`）

改口播轨后 matrix/mg 消费侧自带现场重投影，无需回跑 split。

个性化出路：对垂类玩法/审美本身不满意（想要另一套写稿腔调、字卡风格、节奏观），MUST NOT 在示例图纸上堆个性化补丁——引导用户用 gtrk-style-maker 建他自己的垂类 skill。示例图纸演示的是链路怎么跑，真正个性化的项目工程归用户自建的 skill 持有；示例对它差着火候是定位使然，这句丑话开工时就说。

八、交付与出片

成片出口 = 客户端（OpenCut Gitruck Edition「打开工程」选 gtrk/project.gtrk）：多轨合成（B-roll+字卡+黑底+双音轨混音）只有客户端出片链完整支持；
gtrk render 仅是主轨快照预览（★ 打样误用教训）：只渲一条视频轨、不合成字卡 overlay——可以当粗查混音与 B-roll 落位的预览片，MUST NOT 当成片交付；
云渲必拒：含本地 B-roll 的工程提交云渲会被直接拒绝（local_broll_cloud_render_rejected）——出片走客户端本地导出或 gtrk render 预览，如实告知用户；
交付话术：报产物目录+客户端打开指引+各环节计费实耗；候选轨小眼睛切换、字卡可删可改、BGM 音量客户端可终调、字幕客户端可整轨换样式——把悔棋通道讲清楚。
- 报计费实耗 SHALL 转述 --json 的 credits_estimated（实耗口径，豁免时为 0），MUST NOT 转述 credits_would_be（原价）；exempt_probe:"failed" 时须注明『身份没探到，此数按非豁免保守估』。

九、计费速查

环节	计费	备注
transcript 转写	ASR 按分钟	二创场景一次
matrix index（embed）	会话计量按实际用量	同合云内部成员豁免
matrix describe	1 积分/张	内部成员豁免；同帧缓存永不重复计费
B-roll 云端编排（`matrix lay`，本地素材路）	按「编排量」计费（服务端复算为准）	缺省即走云端（`--arrange local` 已不受理）；素材矩阵路的编排在本机跑、不计费。预估用 `--arrange-estimate-only`（§四.6），跑前须已拿到同意
TTS 配音	按成本档 1-12 积分/计费分钟（240 字/分钟折算）	计费分钟按标准语速折算，不等于音频实际时长；音色 catalog 可查档位；同参重合成走缓存不重复扣费
project init / split / audio lay / subtitle lay	0 积分	`matrix lay` 不在此列（见上一行的云端编排）；producer 同步口 0 元留痕；beat 分析（--beat-align）走 audio_music_analyze 计费
BGM 矩阵搜索（`gtrk matrix material`）	公开口 1 积分/次；矩阵成员口 0 积分	档位由零件自动路由，命令输出如实报口径与计费
image_move 图片运镜	2 积分/张	素材夹含图片且被选中上轨时

★ 脚注（划界）：表里那个 240 字/分钟是计费折算常量， MUST NOT 用于写稿篇幅估算 —— 写稿口径在 §三（挂钟语速，按音色取、缺省 ≈330），两者保守方向相反。本脚注只划界，不动该行的积分档位与计量单位（单位口径归 infra refactor-billing-measure-kchar 的 CLI 联动件）。

起跑前把预估总消耗报给用户（检查点①一并确认），结束报实耗。 §一′ 适配闸命中而中止时，已经发生的计费（transcript / index / describe）SHALL 如实报出——MUST NOT 只报一句「已中止」把账藏起来。

十、已知边界与排障

闪帧：VFR 素材的场景切点可能产生闪帧（专项修复中）——用户报闪帧时告知已立案，客户端可手动微调切点避开；
拆分注释投影：客户端暂不显示 beat 注释（调查中）；
字幕：⑪ gtrk subtitle lay 已直接上好（7 样式 × 11 色，客户端可整轨换样式、逐条可编辑）；改稿重跑 TTS 后须重跑 subtitle lay（时码变了）；
describe 的 text_overlay 信号命中多为撞上原片烧录字幕区（片头片尾高发）——裁定换段或保留交客户端；
matrix lay 拒铺（tracks_edited）不是失败：用户在客户端动过候选轨，plan 照产（planReusable:true），提示用户后按其意愿 --replace-all 或保留；
BGM 曲长 < 成片：缺省循环叠满至工程末尾、尾段裁齐（多 clip 首尾相接，命令打 INFO 报段数）；要单次铺才传 --no-loop。

改工程：走 `gtrk patch`，不要裸手改 JSON

跑完之后要微调某个片段（挪位置 / 改时长 / 切开 / 改音量）时，一律用 gtrk patch：

gtrk patch move --project <dir> --clip <clip_id> --to 5.0
gtrk patch trim --project <dir> --clip <clip_id> --out -1s

MUST NOT 直接编辑 .gtrk 的 JSON。 片段时码是两套并存的（clip_st+clip_ed 与 clip_st+duration）：改一份不改另一份，客户端优先读 clip_ed、后端又不强校验它 ⇒ 没人报错，成片却用了陈旧出点。gtrk patch 替你做恒等式同步 + 帧对齐 + 写前全档校验。

参数手册在 AGENT.md 的「元素级编辑：gtrk patch」一节与 README 命令参考。本节只指路不搬手册 —— 搬过来就会长出会漂移的副本。

About this skill

Automatic

Marketing & Content pack

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v8 — 10/26/2026

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

gitruck / cli

skills/gtrk-travel-recap/SKILL.md

View on GitHub ↗

License MITImported 10/26/2026Version v1Bundled 1 file

Illustrative Want to see what gitruck/gtrk-travel-recap 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.

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

gtrk-travel-recap（旅拍解说 · 组合成片图纸）

一、输入契约与双场景识别

一′、素材适配闸（写稿前必过）

二、全链总览（快速成片）

三、三段式写稿规约

通用框架层（可移植——其他垂类图纸原样复用本层）

旅拍要素层（垂类专属）

三′、L1 看点 rubric · 画面投影版（供 --highlight-weight 使用）

四、检查点①（双模式共有，必停）

五、快速成片编排（打样验证过的真实命令）

六、MG 轻量字卡规约（旅拍向最小集，不建栏目 skill）

七、增量调整与个性化出路（垂类示例定位）

八、交付与出片

九、计费速查

十、已知边界与排障

改工程：走 gtrk patch，不要裸手改 JSON

SkillSafety

Source

Other skills for the same job

Related skills

More in the registry

三′、L1 看点 rubric · 画面投影版（供 `--highlight-weight` 使用）

改工程：走 `gtrk patch`，不要裸手改 JSON