Ready to supercharge your AI agents?

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

Context Surfing · pskoett | DecimalAI

Agent Toolingby pskoett

Context Surfing

Monitors context window health during large, long-running, multi-session, or explicitly context-sensitive work. Uses bounded external anchors, one cold-context review when self-recovery is uncertain, and a clean handoff on real drift. Do not auto-activate merely because a medium task has an intent frame and plan.

Never made a case worseMaintained

SkillScore

SkillScore95/100

Test results 95

Safety

3/3 safety checks passed

Code scan
Intent
Content

What it changes

14%95%of test cases passed +82 pts

Without skill3 of 22

With skill21 of 22

18 fixed3 passed either way1 still failing0 broke

Measured on gemini-3.8-flash11/2/2026v1

Track record

Agent runs: 0; All time

See one task, with and without the skill

One test taskWe need to fix a small spelling typo in the error message string inside auth_helper.ts. Should we activate context-surfing with pre-commit anchor verification and background telemetry tracking for this task?

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 pskoett-context-surfing
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("pskoett-context-surfing")
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 pskoett-context-surfing

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/pskoett-context-surfing/SKILL.md https://app.decimal.ai/s/pskoett-context-surfing/SKILL.md

Copy SKILL.md

Save it in your project, as

.claude/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

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

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
from agents import Agent, Runner

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
from openai import OpenAI

client = OpenAI()
skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
import anthropic

client = anthropic.Anthropic()
skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/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/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
from langchain.agents import create_agent

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

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

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
from pydantic_ai import Agent

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
from deepagents import create_deep_agent

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from agent_framework import Agent

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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("pskoett-context-surfing")
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/pskoett-context-surfing/SKILL.md

Give it to your agent

from pathlib import Path
from strands import Agent

skill = Path(".agents/skills/pskoett-context-surfing/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 pskoett-context-surfing

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/pskoett-context-surfing/SKILL.md https://app.decimal.ai/s/pskoett-context-surfing/SKILL.md

Copy SKILL.md

Save it in your project, as

.agents/skills/pskoett-context-surfing/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/pskoett-context-surfing/SKILL.md

Read the summary as JSON

curl -fsSL https://app.decimal.ai/s/pskoett-context-surfing.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 11/2/2026
Source: pskoett/pskoett-ai-skills ↗
License: Not recorded

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

Context Surfing

Install

gh skill install pskoett/pskoett-skills context-surfing

Fallback using the Agent Skills CLI:

npx skills add pskoett/pskoett-skills/skills/context-surfing

The agent rides the wave of peak context. When the wave crests, it commits. When it detects drift, it pulls out cleanly — saving state, handing off, and letting the next session catch the next wave.

No wipeouts. No zombie sessions. Only intentional, high-fidelity execution.

Mental Model

Think of context like an ocean wave:

Paddling in = loading the intent frame, plan, and initial context. Energy is building.
The peak = full context coherence. The agent knows exactly what it's doing and why. This is when to execute.
The shoulder = context starting to flatten. Still rideable, but output density is dropping.
The close-out = drift. Contradiction, hedging, second-guessing, or hallucinated details. Wipe-out territory.

The skill's job: ride as long as the wave is good, exit before it closes out.

Lifecycle Position

[plan-interview (optional)] → [initial intent frame] → [session-context]
  → [finalize approach] → [context-surfing ACTIVE during execution]
  → [verify-gate] → [simplify-and-harden] → [self-improvement]

Context Surfing is an optional execution layer for work with meaningful context pressure. It wraps execution between intent capture and post-completion review only when the task is Large, Long-running, multi-session, explicitly requests context monitoring, or shows concrete drift risk.

Relationship with intent-framed-agent

When context-surfing is activated, both skills are live during execution. They monitor different failure modes:

intent-framed-agent monitors scope drift — am I doing the right thing? It fires structured Intent Checks when work moves outside the stated outcome.
context-surfing monitors context quality drift — am I still capable of doing it well? It fires when the agent's own coherence degrades (hallucination, contradiction, hedging).

They are complementary, not redundant. An agent can be perfectly on-scope while its context quality degrades (e.g., it's doing the right thing but starting to hallucinate details). Conversely, scope drift can happen with perfect context quality (the agent deliberately chases a tangent). Intent-framed-agent's Intent Checks continue firing alongside context-surfing's wave monitoring.

Precedence rule: If both skills fire simultaneously (an Intent Check and a drift exit at the same time), context-surfing's exit takes precedence. Degraded context makes scope checks unreliable — resolve the context issue first, then resume scope monitoring in the next session.

Cadence separation: Intent checks fire at scope boundaries — "before touching a new area/file, before starting a new logical work unit, when current action feels tangential" (intent-framed-agent/SKILL.md). Context-surfing's pre-commit anchor check fires at side-effecting-action moments — specific tool calls, writes, commits, commit-level output. Don't run both in the same beat: if an Intent Check has just fired and resolved cleanly, the next side-effecting action inside that same work unit doesn't need a fresh anchor check — you already re-grounded.

When to Use the Full Pipeline

Not every task needs every skill in the pipeline. Match pipeline depth to task complexity:

Task Type	Skills to Use
Trivial (rename, typo fix)	None — just do it
Small (isolated bug fix, single-file change)	`verify-gate` + `simplify-and-harden`
Medium (feature in known area, multi-file)	`intent-framed-agent` + `session-context` + `verify-gate` + `simplify-and-harden`
Large (complex refactor, new architecture, unfamiliar codebase)	Full pipeline
Long-running (multi-session, high context pressure)	Full pipeline with `context-surfing` as the critical skill

This table matches skill-pipeline's task classification, which is the canonical routing source.

When in doubt, start light. Add skills if you notice drift or quality issues mid-task.

Activation

Activate this skill when the task is Large or Long-running, work spans sessions, the user explicitly requests context monitoring, or concrete behavior indicates drift risk. An intent frame plus a plan is not sufficient by itself; do not auto-activate for an ordinary Medium task.

At activation, load the smallest useful anchor:

The user's current task and latest scope change
The applicable agent instruction file(s)
The approved intent frame and plan, when present
A prior handoff, when resuming
The current session state from Entire, when available

If neither an intent frame nor a plan exists (standalone mode), use the user's original task description combined with project context files as the wave anchor. This is sufficient — the skill degrades gracefully, not catastrophically.

Entire CLI Integration

Entire CLI (github.com/entireio/cli) provides persistent session state that serves as external ground truth for drift checks and handoff files.

At activation, detect Entire:

entire status 2>/dev/null

If it succeeds and the active host adapter supports the relevant lifecycle events, Entire captures those supported events passively via hooks. Do not infer full transcript or checkpoint coverage from entire status alone. Use entire status and entire explain --session <id> to inspect available state and record unsupported host coverage explicitly.
If unavailable or failing, record the optional integration as unavailable with the bounded reason, then continue without it. Use the intent frame and plan file as the wave anchor instead. Do not turn optional telemetry failure into a delegation or policy denial, and do not repeatedly retry or nag about installation.

The wave anchor is not held mentally. It is built from whatever external, persistent artifacts are available — every drift check reads from them directly, never from reconstructed memory.

Handoffs become learning signals

When context-surfing writes a handoff file on drift exit, a supported host adapter may also capture the related Stop / SessionEnd events in Entire. At cadence, learning-aggregator --deep reads only the transcript coverage that is actually available and extracts:

Which drift signals most commonly trigger handoffs → potential session design issues
How many sessions hit drift exit vs completed cleanly → session health baseline
What context artifacts were missing when drift occurred → promotion candidates for project instruction files

Structured handoff files (.context-surfing/handoff-<slug>-<timestamp>.md) with predictable section headers make this parseable. You don't need to do anything special — just keep the handoff format consistent.

Full pipeline: intent frame + plan file + Entire CLI session state
Partial pipeline: whichever of intent frame or plan exists, plus the applicable instruction file
Standalone: the user's original task description + the applicable instruction file

The anchor is weaker in standalone mode — fewer artifacts to cross-check against — but it is always present. A weak anchor is better than no anchor.

Project Context Files

Before executing, identify the minimum standing context needed for the task. Do not scan or load every root Markdown file by default.

Always load at activation

The applicable agent instruction file(s), such as AGENTS.md, CLAUDE.md, or .github/copilot-instructions.md
The user's current task and latest scope change
The approved plan/intent artifacts, when present
The selected handoff file, when resuming

Load on demand (when relevant to the current task)

README.md for project purpose or structure
A specific root Markdown file named by the active instructions or task
.md files in skills/, docs/, .learnings/, or similar directories
SKILL.md files for any skill being invoked alongside this one

Rules for context files

They are always part of the wave anchor. If output contradicts a project .md file, that is a drift signal — treat it as a strong one.
Re-read them if the task changes domain or scope. Don't assume you remembered correctly 20 steps in.
Include their key constraints in the handoff file. The next session needs to reload them too — note which files were active and whether any were updated during the session.
If a .md file is modified during the session, flag it explicitly in the handoff under a "Modified Context Files" section so the next session re-reads it fresh rather than relying on the handoff summary.

Drift Detection

Continuously monitor for these signals. Strong signals trigger an immediate exit — the wave is closing out. Weak signals trigger the Recovery Protocol first — the shoulder might still be rideable.

These signals apply to your own reasoning chain as much as to your emitted output — scan your thinking, not just your text. Drift originates inside the reasoning trace before it surfaces as visible prose, and the anchor's job is to gate reasoning, not just to audit output after the fact. A reasoning-level check before a side-effecting action is strictly upstream of an output-level check afterwards.

Strong signals (exit immediately)

The agent contradicts a decision it already made and committed to
The agent commits to a detail contradicted by the anchor and cannot reconcile it
The agent re-opens a scope question that was explicitly resolved in the plan

Weak signals (trigger recovery)

Responses are getting longer without getting more useful
Hedging language increases: "it depends", "could be", "might want to consider"
The agent switches approaches mid-task without explicit user direction
References to the original intent become vague or paraphrased instead of precise
The agent asks a clarifying question it should already know the answer to
Output starts re-explaining the task rather than executing it
A new unsupported detail appears and needs verification against the anchor

Not drift

Normal iteration and refinement within scope
Asking about genuinely new information not in the original context
Simplifying a previous output (that's the wave working, not breaking)

The Monitoring Paradox

An agent with degraded context is the least likely to detect its own degradation. The strong signals (hallucination, contradiction) are exactly the ones a compromised agent will miss — because it no longer has the context to recognize the contradiction.

This is an inherent limitation of self-monitoring. Mitigate it with external grounding:

Pre-commit anchor check. Before any side-effecting action (file write, commit, external tool call that changes state, user-visible commit-level output), run this inline in your reasoning chain:
- Quote the relevant anchor line verbatim from a fresh read — not from recall. If an intent frame exists, open and quote from it. If a plan file exists, quote the relevant section. In standalone mode, quote from the user's original task description.
- State the pending action in one line.
- Confirm the action traces back to the quoted line.
- If traceability is unclear, do not just "stop" — enter the Recovery Protocol below. "Stop" without a named next branch defaults to whichever shape is cheapest, which is usually silently continuing.
Anchor mutability: If a context file has been modified mid-session, the pre-commit check uses the latest version of the file, not the activation-time snapshot. Re-read cold.
Use Entire CLI logs as external memory. If Entire is available, read back your own session log before non-trivial decisions. Your logged state is more reliable than your recalled state.
Treat the user as a drift sensor. If the user expresses confusion, asks "why are you doing that?", or redirects you — treat it as a strong signal regardless of your own assessment.

The weak signals (hedging, verbosity) are more reliably self-detectable precisely because they're behavioral, not factual. Watch for those as early warnings.

Recovery Protocol (Wave Re-Anchor)

When weak signals accumulate, don't exit immediately — try to re-anchor first. The shoulder of the wave is still rideable if you can re-ground.

Step 1: Pause and re-read

Stop producing output. Re-read whatever wave anchor artifacts are available:

If an intent frame exists, open and read it verbatim (not your memory of it)
If a plan file exists, open and read the relevant section
In standalone mode, re-read the user's original task description and project context files
If Entire CLI is available, read back your session log

Compare what you're currently doing against what these artifacts say you should be doing.

Step 2: Reconcile

If the mismatch resolves — you can trace a clear line from the anchor to your current work — resume execution. The weak signals were noise, not drift.
If uncertainty remains, use one general-purpose read-only subagent for a cold-context second opinion before escalating. Supply the intent frame, plan paths or original task, current next action, and observed signals. Ask it to read the anchors and applicable project instructions directly, compare the work with committed decisions, and check the strong and weak drift signals above. Require a report with status (healthy / weak signals / strong drift), anchors read, concrete observations, alignment, and a recommendation to continue, re-anchor, or hand off. Use read-only tools where supported and prohibit edits in the brief. In Copilot, request the same review in chat. If independent execution is unavailable, disclose that limit and escalate rather than claiming a cold-context check.
If the reviewer returns "healthy" or "weak signals only", integrate its corrections and resume execution.
If the reviewer returns "strong drift detected", or you still can't reconcile after its report, escalate to the user. Explain the mismatch and the review findings, then ask how to proceed.
If the user re-grounds you — integrate their clarification and resume. The re-anchor succeeded.
If the user can't resolve it, or the original intent has fundamentally changed — proceed to the Exit Protocol.

Re-anchor limits

Allow at most one local re-anchor and one cold-context check for the same drift episode. If they do not resolve it, escalate or exit. Strong signals bypass recovery and trigger an immediate exit.

Riding the Wave

While context is healthy:

Execute with commitment. No hedge, no re-litigating decisions already made. The plan is the plan.
Run the pre-commit anchor check before side-effecting actions. See the Monitoring Paradox section for the protocol (quote verbatim → state pending action → verify trace-back → on failure enter Recovery Protocol, don't just "stop"). This check is the primary context-quality gate during execution — not a nice-to-have. Scope-level checks are intent-framed-agent's job and fire at different boundaries (see Relationship with intent-framed-agent for cadence separation).
Track completions. Log what's done, what's in progress, what's pending as work progresses — not at exit. If Entire CLI is available, use it as the session log. If not, maintain a running status in your output. This feeds the handoff if needed.
Resist scope creep. If something interesting but out-of-scope appears, note it (in Entire CLI or in your output) — don't chase it.

Exit Protocol (Wave Close-Out)

When a strong signal fires, or the Recovery Protocol fails to re-anchor, execute a clean exit. Do not try to push through.

Step 1: Stop executing

Immediately pause task execution. Do not produce more output that may be corrupted by the degraded context.

Step 2: Write the handoff file

Create a file named .context-surfing/handoff-[slug]-[timestamp].md (create the .context-surfing/ directory if it doesn't exist). Add .context-surfing/ to .gitignore — handoff files are session artifacts, not project history.

Structure:

# Context Surf Handoff

## Session Info
- Task: [task name / slug]
- Started: [timestamp]
- Ended: [timestamp]
- Exit reason: [what drift signal was detected]

## Intent Frame (if available — read verbatim, do not reconstruct)
[copy directly from the intent-framed-agent artifact, or "N/A — standalone session" if none exists]

## Plan (if available — read verbatim, do not reconstruct)
[copy directly from the plan-interview artifact, or "N/A — standalone session" if none exists]

## Original Task Description (standalone fallback)
[copy the user's original task description if no intent frame or plan exists]

## Completed Work (from Entire CLI session state or running output log)
[pull directly from CLI log or structured output — do not reconstruct from memory]

## In Progress at Exit (from Entire CLI session state or running output log)
[what the session log shows as active at the moment of exit]
[include any partial outputs if useful]

## Pending Work (from plan-interview output — cross-reference session log to confirm what's genuinely not done)
[remaining tasks from the plan, in order]

## Drift Notes
[what specifically triggered the exit — be precise, this helps the next session replan intelligently]

## Active Context Files
[list all .md files loaded during this session — root level and any skill/docs files consulted]

## Modified Context Files
[any .md files that were changed during this session — next session must re-read these, not trust the handoff summary]

## Scope Notes (Out-of-Band)
[anything interesting discovered that's outside scope — flagged for the next session to decide on]

## Recommended Re-entry Point
[where the next session should pick up, and any replanning it should do before continuing]

Step 3: Notify the user

Briefly and clearly:

"Context wave is done. I've saved the session state to .context-surfing/[filename]. The next session should load that file and resume from [re-entry point]. Replan only if the handoff records unresolved requirements or a material scope change. Here's what triggered the exit: [one sentence on drift signal]."

Do not over-explain. The handoff file has the details.

Handoff to Next Session

The next session should:

Read the handoff file completely before doing anything else
Reuse an approved plan and intent frame when the handoff shows they remain complete and unchanged
Run plan-interview only when the handoff records unresolved requirements, a material scope change, or an invalidated plan
In standalone mode, re-ground from the task description and drift notes
Resume from the recommended re-entry point
Mark the handoff consumed only after re-entry succeeds

This is not failure. This is the system working correctly. Clean exits produce better total output than zombie sessions that limp to the finish line.

Session Close (Natural Completion)

When the task completes within a healthy wave (no drift exit needed):

Confirm all plan items are done (or, in standalone mode, confirm the original task description is satisfied)
Note session end in a brief summary (optional, not a full handoff file)
Signal readiness for simplify-and-harden — the next skill in the pipeline picks up from here

No handoff file needed for clean completions — just the outputs and a one-liner status.

Interoperability with Other Skills

What this skill consumes

From plan-interview (optional): The plan file (docs/plans/plan-NNN-<slug>.md). Strengthens the wave anchor when available.
From intent-framed-agent (optional): The intent frame artifact. Strengthens the wave anchor when available.
From the user (always available): The original task description. Used as the standalone wave anchor when upstream artifacts are absent.
From Entire CLI (optional): Session state for progress tracking and external memory.

What this skill produces

For simplify-and-harden: A "ready" signal on natural completion. Simplify-and-harden picks up from the completed work.
For the next session (on drift exit): A handoff file in .context-surfing/ with full state for session resumption.
For self-improvement: Drift patterns observed during the session can be logged as learnings if they recur.

Pipeline position

plan-interview (optional, for requirement shaping)
intent-framed-agent (initial execution contract + scope drift monitoring)
session-context (prior-session evidence before finalizing the approach)
context-surfing (optional context-quality monitoring for Large, Long-running, or explicitly context-sensitive execution)
verify-gate (machine verification: compile + test + lint)
simplify-and-harden (post-completion quality/security pass)
self-improvement (capture recurring patterns and promote durable rules)

Hook Integration

Enable automatic handoff detection at session start. This ensures handoff files from previous context exits are never silently ignored.

Setup (Claude Code)

Add to .claude/settings.json. Point the command at where the skill is installed (.claude/skills/context-surfing/ for gh skill install / npx skills add, skills/context-surfing/ if this repo is vendored; the plugin bundle registers this hook automatically with a ${CLAUDE_PLUGIN_ROOT} path):

{
  "hooks": {
    "SessionStart": [{
      "hooks": [{
        "type": "command",
        "command": "${CLAUDE_PROJECT_DIR}/.claude/skills/context-surfing/scripts/handoff-checker.sh"
      }]
    }]
  }
}

Codex CLI supports the same SessionStart event via <repo>/.codex/hooks.json (experimental, behind codex_hooks = true in config.toml); plain stdout from the script is added as developer context there too.

This checks for pending handoff files in .context-surfing/ at session start. After a successful re-entry, create the adjacent handoff-<slug>-<timestamp>.md.consumed marker so future sessions do not repeat the same reminder. The marker is session state inside the already ignored .context-surfing/ directory.

Copilot / Chat Fallback

For agents without hook support, manually check at session start:

ls .context-surfing/handoff-*.md 2>/dev/null

Principles

Ride the peak, not the whole ocean. A shorter session with high fidelity beats a long session with gradual corruption.

Exit is not failure. The wave close-out is a feature. Detecting it early is the skill.

The handoff file is the continuity. It's not documentation overhead — it's what makes the next session as sharp as this one started.

Never hide the exit. Always be explicit with the user that a context exit happened and why. Silently continuing in degraded context is the worst outcome.

About this skill

Automatic

AI Engineer 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

pskoett / pskoett-ai-skills

skills/context-surfing/SKILL.md

View on GitHub ↗

Imported 10/26/2026Version v1Bundled 1 file · scripts

Illustrative Want to see what pskoett/context-surfing would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

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

Context Surfing

Install

Mental Model

Lifecycle Position

Relationship with intent-framed-agent

When to Use the Full Pipeline

Activation

Entire CLI Integration

Handoffs become learning signals

Project Context Files

Always load at activation

Load on demand (when relevant to the current task)

Rules for context files

Drift Detection

Strong signals (exit immediately)

Weak signals (trigger recovery)

Not drift

The Monitoring Paradox

Recovery Protocol (Wave Re-Anchor)

Step 1: Pause and re-read

Step 2: Reconcile

Re-anchor limits

Riding the Wave

Exit Protocol (Wave Close-Out)

Step 1: Stop executing

Step 2: Write the handoff file

Step 3: Notify the user

Handoff to Next Session

Session Close (Natural Completion)

Interoperability with Other Skills

What this skill consumes

What this skill produces

Pipeline position

Hook Integration

Setup (Claude Code)

Copilot / Chat Fallback

Principles

SkillSafety

Source

Related skills

More in the registry