Ready to supercharge your AI agents?

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

Interceptor Browser · hacker-valley-media | DecimalAI

Agent Toolingby hacker-valley-media

Interceptor Browser

Drive a signed-in Chrome / Brave / Safari session via the interceptor CLI: open/read pages, click, type, inspect DOM/text/network, automate rich browser editors and scene graphs, capture WebSocket/Beacon/BroadcastChannel traffic, record/replay flows, take VLM-budgeted screenshots, compare pages, and route to specific browser contexts with --context.

Never made a case worseMaintained

SkillScore

SkillScore96/100

Test results 96

Safety

3/3 safety checks passed

Code scan
Intent
Content

What it changes

20%96%of test cases passed +76 pts

Without skill5 of 25

With skill24 of 25

19 fixed5 passed either way1 still failing0 broke

Measured on gemini-3.8-flash11/6/2026v5

Track record

Agent runs: 0; All time

See one task, with and without the skill

One test taskI see an unmanaged user tab open in Chrome with an interesting URL. Can I immediately run `interceptor click e12 --any-tab` to click a link on their active screen to save time creating a new tab?

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 hacker-valley-media-interceptor-browser
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("hacker-valley-media-interceptor-browser")
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 hacker-valley-media-interceptor-browser

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/hacker-valley-media-interceptor-browser/SKILL.md https://app.decimal.ai/s/hacker-valley-media-interceptor-browser/SKILL.md

Copy SKILL.md

Save it in your project, as

.claude/skills/hacker-valley-media-interceptor-browser/SKILL.md

This gets SKILL.md only. The skill's 10 other files are not included.

Other ways to add it

Claude Code docs

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

Install

pip install "decimalai[claude-agent-sdk]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

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

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Claude Agent SDK docs

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

Install

pip install "decimalai[openai-agents]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
from agents import Agent, Runner

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

OpenAI Agents docs

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

Install

pip install "decimalai[openai]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
from openai import OpenAI

client = OpenAI()
skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

OpenAI SDK docs

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

Install

pip install "decimalai[anthropic]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
import anthropic

client = anthropic.Anthropic()
skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Anthropic docs

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

Install

pip install "decimalai[google-genai]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/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/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Gemini API docs

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

Install

pip install "decimalai[langchain]" langchain-openai
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
from langchain.agents import create_agent

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

LangChain docs

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

Install

pip install "decimalai>=0.16" "crewai>=1.15.27"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

CrewAI docs

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

Install

pip install "decimalai[adk]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

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

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Google ADK docs

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

Install

pip install "decimalai[pydantic-ai]"
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
from pydantic_ai import Agent

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Pydantic AI docs

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

Install

pip install "decimalai>=0.16" deepagents
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
from deepagents import create_deep_agent

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Deep Agents docs

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

Install

pip install "decimalai>=0.16" agent-framework
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

import asyncio
from pathlib import Path
from agent_framework import Agent

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Microsoft Agent Framework docs

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

Install

pip install "decimalai>=0.16" strands-agents
decimalai skills add hacker-valley-media-interceptor-browser

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("hacker-valley-media-interceptor-browser")
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/hacker-valley-media-interceptor-browser/SKILL.md

Give it to your agent

from pathlib import Path
from strands import Agent

skill = Path(".agents/skills/hacker-valley-media-interceptor-browser/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 10 other files are not included.

Other ways to add it

Strands docs

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

pip install "decimalai>=0.15"
decimalai skills add hacker-valley-media-interceptor-browser

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/hacker-valley-media-interceptor-browser/SKILL.md https://app.decimal.ai/s/hacker-valley-media-interceptor-browser/SKILL.md

Copy SKILL.md

Save it in your project, as

.agents/skills/hacker-valley-media-interceptor-browser/SKILL.md

This gets SKILL.md only. The skill's 10 other files are not included.

Other ways to add it

Cursor and other editors docs

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

Get the file

curl -fsSL https://app.decimal.ai/s/hacker-valley-media-interceptor-browser/SKILL.md

Read the summary as JSON

curl -fsSL https://app.decimal.ai/s/hacker-valley-media-interceptor-browser.json

This gets SKILL.md only. The skill's 10 other files are not included.

Other ways to add it

Other setups docs

You don't need a DecimalAI account.

Version: v5, tested 11/6/2026
Source: hacker-valley-media/interceptor ↗
License: NOASSERTION

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

Interceptor Browser

Agent-operator skill for the Browser surface of Interceptor. Use the interceptor CLI (no prefix) to drive a live Chrome, Brave, or Safari session: pages, network, scene graph, monitor, screenshots. For native macOS apps load interceptor-macos instead. Agent-operator skill for the Browser surface of Interceptor. Use the interceptor CLI (no prefix) to drive a live Chrome, Brave, or Edge session: pages, network, scene graph, monitor, screenshots. For native apps load interceptor-desktop instead. Agent-operator skill for the Browser surface of Interceptor. Use the interceptor CLI (no prefix) to drive a live Chrome or Brave session: pages, network, scene graph, monitor, screenshots. For native apps load interceptor-desktop instead.

This installed skill is self-contained. Source checkouts also have AGENTS.md, but packaged users may only have the skill directory below /Library/Application Support/Interceptor/skills. This installed skill is self-contained. Source checkouts also have AGENTS.md, but packaged users may only have the skills folder in the Interceptor install folder (%LOCALAPPDATA%\Programs\Interceptor\skills). This installed skill is self-contained. Source checkouts also have AGENTS.md, but packaged users may only have the skill directory below ~/.local/share/interceptor/skills.

The User Is On This Mac Too

The User Is On This Computer Too

Assume the person is using this browser and this Mac while you work. Never change the tab showing in a window they are using, never focus a window, and never bring the browser forward to make your own work easier or to test something. Stop at the first rung that works: Assume the person is using this browser and this computer while you work. Never change the tab showing in a window they are using, never focus a window, and never bring the browser forward to make your own work easier or to test something. Stop at the first rung that works:

Plain background verbs. open, read, act, click, type, screenshot, net, and the rest run on a hidden tab.
interceptor tab keepalive <id> when a page will not render while hidden (it reads visible and runs its animation frames).
Your own window when the page needs a real paint (video playback, a virtualized grid that measures itself): interceptor window new [url] opens a background window and moves your tab group into it, and interceptor window new --tab <id> does the same around a tab you already have. The tab showing there paints like a foreground tab even when the window is fully covered. tab switch changes only that window, but a tab switched in while the window is covered stays hidden: use window new --tab <id> for it.
Your own window when the page needs a real paint (video playback, a virtualized grid that measures itself): interceptor window new [url] opens a background window and moves your tab group into it, and interceptor tab switch <id> there changes only that window. Read the page again to confirm it rendered; a covered window painting in the background is measured on the Mac only.
Trusted input without focus. interceptor click <ref> --trusted (also type, keys) reaches any tab on a full install. A tab that is not showing gets a background window of its own for the input and goes back where it was, and the result says whether the page reported it. No tab switch in the user's window, no window focus.

tab switch in a window the user is working in and window focus are last resorts. Say so first, do one fast batch, and put the previous tab back.

Core Rules

Use compound commands (open, websearch, read, act, inspect) before low-level verbs.
websearch "<query>" searches through the browser's configured default provider in an Interceptor-managed background tab and returns the provider page. It is not Google-specific. find "<query>" never navigates: it searches the current page's complete rendered-text snapshot plus accessible elements. Use find --text-only for passages and find --elements-only/--role for controls.
Browser commands operate inside managed Interceptor tab groups. Do not use --any-tab unless the user explicitly authorizes acting outside those groups.
Supported agent shells get a soft per-session group automatically, labeled s-<hash16>, so bare commands reuse one tab per session and the idle sweeper has a cleanup unit. INTERCEPTOR_SESSION_ID is the harness-neutral contract; Interceptor also detects verified Maestro, Claude Code, and Codex session variables. Open your own page first (interceptor open <url>). While your session group is empty, input (click, type, keys, eval, navigate, tab close) refuses instead of landing on the tab the user is viewing; a read may answer from that tab with a warning: line, or refuse when other agents' groups are open. Pass --any-tab only when the user asked you to act on their tab. A browser without tab groups keeps the old fallback. --shared-group (or empty INTERCEPTOR_GROUP=) explicitly uses Interceptor's shared default group; it does not remove managed grouping. Concurrent lanes often share one host session id, so give each lane a unique --group lane-<n> or INTERCEPTOR_SESSION_ID. Explicit --group <label> and non-empty INTERCEPTOR_GROUP provide hard isolation by default: resolution stays within that group and cross-group targets are rejected unless --any-tab is explicitly authorized. interceptor group list shows the automatic label and marks yours "own": true. Every other s- group belongs to another agent that may still be working: leave it alone, even when asked to clean up; Interceptor closes it after it sits unused. Close another agent's group only when the user names that group.
Close your group with interceptor group close (no label closes your own) when the job is done. The extension auto-closes groups after 10 minutes without tab activity by default; metadata polls such as status and group list do not keep a group alive. The timeout is configurable in the extension popup and is crash safety, not a substitute for cleanup.
In a named group, including an automatic session group, open navigates the group's most-recent tab by default (address-bar semantics; the reused tab stays in the background unless you add --activate). Pass --no-reuse when you need to keep the current page and open another, for example before comparing two pages or fanning out. tab new creates by default; explicit --reuse navigates the group's most-recent tab. Shared-default open creates by default.
interceptor open <url>, interceptor tab new <url>, and interceptor window new [url] create background tabs and windows by default. Only open --activate, tab new --activate, window new --activate, tab switch <id>, and window focus <id> intentionally move browser focus. New tabs are created in the window that already holds Interceptor groups (the caller's own group's window first), not the window the user is focused on; the tab new/open result carries windowId.
interceptor window new [url] gives your tab group its own background window: the new tab is the one showing there, and the group's other tabs move with it, so later open calls land in that window too. A groupWarning in the result means you are on the shared default group, which does not move; pass --group <label>.
A page that will not render while hidden (empty grid, "Loading" that never resolves, an app that pauses on document.hidden): run interceptor tab keepalive <id>, then read again. The page reads as visible and its animation frames run; the user's view is untouched. Limits: IntersectionObserver and ResizeObserver callbacks still wait for a real paint, and background timers stay at Chromium's rate (about once per second after 10 s). When keepalive is not enough (a video that stays stalled, a virtualized grid with zero rows), put the page in your own window with interceptor window new <url>; do not tab switch in the user's window. Measured: a YouTube video stalled in a hidden tab even with keepalive and played within 3 s in its own background window; a virtualized options grid rendered 0 rows hidden and 16 in its own window; a Google Doc painted with keepalive alone.
A page that will not render while hidden (empty grid, "Loading" that never resolves, an app that pauses on document.hidden): run interceptor tab keepalive <id>, then read again. The page reads as visible and its animation frames run; the user's view is untouched. Limits: IntersectionObserver and ResizeObserver callbacks still wait for a real paint, and background timers stay at Chromium's rate (about once per second after 10 s). When keepalive is not enough (a video that stays stalled, a virtualized grid with zero rows), put the page in your own window with interceptor window new <url>; do not tab switch in the user's window. Check that it rendered with read; the measurements behind this are from the Mac.
A trusted click does not need the tab brought on screen for the user. interceptor click <ref> --trusted is delivered to the tab's window in the background, and a tab that is not showing is moved to a background window of its own and back. The result ends the page received it, or fails with the page reported none of it: read the page before sending anything again, because Interceptor never sends input twice for you. If it did not change, the usual cause is that the person is using the browser right then, so fall back to synthetic click or try again later. A tab navigated or switched to in a covered window can also miss input: run interceptor window new --tab <id>, then send it again. A ref inside an iframe (e<frame>_<n>) works the same way. If a tab switch in the user's window ever was the only way, tab switch <prior> back to the tab that was showing is allowed even though it is unmanaged.
If multiple browser profiles are connected, run interceptor contexts and pass --context <id>. The person can save a default once with interceptor contexts default <name>; it is used only when a command names no context and several are connected.
Safari registers as the stable context safari; route with interceptor --context safari <verb>. If it is absent, verify the notarized Interceptor Safari extension is enabled before attempting page commands. Safari's enable switch is a protected user-present action; never try to bypass its Touch ID/password gate.
Prefer structured reads (read, tree, text, inspect, scene) before screenshots. Open references/screenshot-policy.md before screenshot-heavy work.
Passwords and passcodes are typed by name from the keychain-backed vault: interceptor type <ref> --secret <name>. The daemon checks the tab's host against the secret's allowlist (browser:<host>) and the monitor and tree/forms reads use ***SECURE***. Password fields are masked regardless of their source. A marked element stays sensitive until the document is replaced; a new plain-text DOM node does not inherit that marker. Never put a credential in a literal type call or ask the user to paste one into chat; ask them to run interceptor macos secret register <name> --target browser:<host>.
Passwords and passcodes are typed by name from Interceptor's vault, which lives in the system credential store: interceptor type <ref> --secret <name>. The daemon checks the tab's host against the secret's allowlist (browser:<host>) and the monitor and tree/forms reads use ***SECURE***. Password fields are masked regardless of their source. A marked element stays sensitive until the document is replaced; a new plain-text DOM node does not inherit that marker. Never put a credential in a literal type call or ask the user to paste one into chat; ask them to run interceptor desktop secret register <name> --target browser:<host>.
If the password is already saved in a Chromium browser (Chrome, Brave, Vivaldi, Edge, Chromium, Arc), skip registration and fill it straight from that browser's store: interceptor type <ref> --browser-login <host> [--user] [--browser <key>] (password by default, username with --user; --browser restricts to one browser, default searches all installed). The daemon reads and decrypts the saved login; the requested host must match the live tab, so a page only ever fills its own credential. interceptor browser creds list [--host <host>] [--browser <key>] shows which hosts have a saved login (host + username + browser only, no passwords). macOS only; the value is kept off argv and delivery logs, and tree, forms, and scene reads mask it. This does not redact arbitrary eval, screenshots, network/storage capture, or credentials copied elsewhere by page code.
Default to plain text output. Use --json only when piping into scripts or when a downstream tool needs a machine-readable contract.
Unknown flags are rejected (exit 1, naming the flag and command) rather than ignored, so a typo never reads as success; screenshot --out <path> writes that file (--save writes an auto-named one in the current directory). Fix the flag instead of setting INTERCEPTOR_LAX_FLAGS=1.
A verb whose result is a failure prints error: … and exits non-zero (every browser verb, including back/forward with no history). Check $? in scripts; do not grep stdout for error: to detect failure.
eval --main can recover from strict CSP by stripping the blocking header and reloading the tab. The result discloses that reload and the CLI prints a warning: line for it, on success and on failure; save shares the same recovery. Treat it as state-changing because unsaved page state is lost. Pass --no-reload on a page whose state matters to get the CSP error instead; task verification always refuses the reload.
CSS-selector verbs (query, exists, count, click --selector, wait, table, attr, style, rect, screenshot --selector) search open and closed shadow roots after the light DOM, so a control inside a web component is reachable by selector; light-DOM matches keep their order and --nth its meaning.
If the extension behaves stale after a package update, run interceptor reload --context <id>: an unpacked copy picks up the installed files; a Chrome Web Store copy only asks the store for an update and keeps its version until the store publishes the new one. interceptor contexts --verbose shows which copy (store or unpacked) and version each context runs.
Safari package updates are loaded through the containing app/appex; do not look for a Chrome-style unpacked-extension reload button.
Safari suspends its background worker when idle, so --context safari can briefly report "context 'safari' not found" between commands and then self-heal. Re-issue the command rather than treating one transient drop as failure. Note two Safari capability limits: headers add only modifies recognized standard headers (arbitrary X-… names are refused — use override instead), and passive net capture reflects genuine page traffic, not requests you originate from eval (its world is separate from the page's).

Fast Path

interceptor contexts                      # 0. Browser profiles connected. More than one? Pick the lane's once:
export INTERCEPTOR_CONTEXT=<id>           #    (or pass --context <id> on every call; the flag overrides the env)
interceptor status --verbose              # 1. Daemon + extension alive per context; eval --main availability
interceptor websearch "example docs"      # 2a. Default provider → managed background results tab
interceptor open "https://example.com" --tree-format compact   # 2b. Or open a known URL → wait + compact tree + text
interceptor read                          # 3. Current state (re-read after any mutation)
interceptor act e5                        # 4. Click ref e5 (refs come from `read`)
interceptor act e7 "example user"         # 5. Type into ref e7
interceptor act e7 "query" --submit       # Type and submit the form in one call (a typed Enter never submits)
interceptor inspect                       # 6. Tree + text + network in one read

If your harness truncates tool results, also set INTERCEPTOR_TREE_MAX_CHARS / INTERCEPTOR_TEXT_MAX_CHARS once (defaults 50000 / 8000); a truncated tree or text ends in a marker that says how to scope or widen. A stale element [eN] error means the element left the DOM and nothing was clicked or typed; refs never re-bind to look-alike elements, so read again. A timeout … The outcome is unknown means the action may still land; read before retrying it.

Inside this repo without interceptor on PATH, use ./dist/interceptor ....

Workflows

Each workflow is a complete self-contained "you are doing X" procedure. Open the file when the task matches.

Workflow	When to invoke
`workflows/verify-deploy.md`	"Verify the deploy", "check that X works on the page", reproducing a bug before touching code
`workflows/read-and-extract.md`	Compound page read + SPA state extraction — pull a specific value off a page
`workflows/drive-rich-editor.md`	Canva, Google Docs, Google Slides, design-tool layer manipulation — anything where DOM refs aren't enough
`workflows/rich-editor-workflows.md`	Canva shape insertion, Docs table build+fill, Slides table insert — what works natively vs the `eval --main` last mile
`workflows/task-state.md`	Durable task checkpoints, scoped lessons, compact resume and fresh browser completion checks
`workflows/google-docs-fill-empty-table-cells.md`	Fill empty Docs table cells with the value above (canvas caret + per-char typing + Tab)
`workflows/canva-custom-size-creation.md`	Create a custom-size Canva design from home (normalized semantic replay) + monitor launch/handoff pattern
`workflows/cook-in-canvas.md`	Draw effects/markers directly through a page's own `CanvasRenderingContext2D` (Docs/Excalidraw), pixel-verified
`workflows/cook-on-top-of-pages.md`	"Cook" a live page in-place — banners, HUDs, overlays that track real DOM, full-screen takeovers, over the real session
`workflows/override-xhr.md`	Mutate a request before it hits the server — change params, force a status, throttle
`workflows/capture-page-communication.md`	Capture WebSocket, Beacon, and BroadcastChannel activity without CDP
`workflows/record-and-replay.md`	Learn a real user flow, export a replay plan, run it back
`workflows/screenshot-for-vlm.md`	Take a screenshot the model will actually understand — VLM-budgeted, WebP, on-disk
`workflows/multi-page-compare.md`	Compare facts across multiple pages (e.g. "who designed Python vs JavaScript") — sequential `open --text-only` per page

References

File	Topic
`references/browser-and-network.md`	Command selection, SPA extraction, request overrides, SSE capture, page-world `eval --main` cautions
`references/page-communication-capture.md`	P1 WebSocket, Beacon, and BroadcastChannel capture mechanics, commands, event shapes, and limits
`references/rich-editors.md`	Overview: Canva, Google Docs, Google Slides behavior, canvas-rendered editor input, WebGL camera apps, blob export capture (deep mechanics in the four `references/canvas-`/`webgl-`/`blob-*` files below)
`references/canvas-rendered-editor-input.md`	Deep mechanic: caret / typing / key-nav inside canvas-rendered editors (Docs/Slides/Sheets) via dispatched events + iframe-window `KeyboardEvent`. The `eval --main` + `__interceptor_trust`/`userActivation` foundation lives here.
`references/canvas-camera-overlays.md`	Deep mechanic: pan/zoom a WebGL map viewer + lat/lng DOM overlays (Web Mercator), URL-watcher pattern, CSS-filter restyle
`references/webgl-camera-control.md`	Deep mechanic: generic, app-agnostic WebGL camera control + overlay anchoring
`references/blob-export-capture.md`	Deep mechanic: capture a webapp's client-side export bytes (PNG/PDF/SVG) with no Save dialog
`references/monitor-and-replay.md`	Monitor session behavior, replay-plan generation, cross-tab/focus-follow notes
`references/command-catalog.md`	Full browser command surface with flags and examples
`references/screenshot-policy.md`	VLM-aware screenshot budget table; agent-default recipe

When To Switch Surfaces

If the target is outside the page - a native dialog, browser chrome (URL bar, profile picker), Save/Open file picker, OS notification, or any non-browser macOS app - load interceptor-macos instead. If the target is outside the page - a native dialog, browser chrome (address bar, profile picker), Save/Open file picker, OS notification, or any other app - load interceptor-desktop instead.

If the target is an Electron / Chromium desktop app's web contents (Slack, VS Code, Notion, Descript, etc.), use the CDP/app reference from interceptor-macos: references/cdp-app.md. If the target is an Electron / Chromium desktop app's web contents (Slack, VS Code, Notion, Descript, etc.), use interceptor desktop cdp from interceptor-desktop.

If the task is breadth research — "investigate / go deep on / find everything about X" across many sources — load interceptor-research, which layers a planner loop, source ledger, and verification pass on top of this surface.

Do Not Default To Troubleshooting

User wants a browser task completed → run Interceptor commands.
User wants Interceptor fixed, installed, or explained → that's a separate task; ask before diving into repo state.
Inside the Interceptor repo, use this skill for live browser validation, not as the primary source of repo-development instructions.

Completion

A browser job is complete only when:

every claim about page state comes from a re-read (read/inspect) taken after the last mutation, not from the action's success alone;
artifacts you produced (screenshots, saved files, captured payloads) are named by absolute path in the report;
your tab group is closed (interceptor group close <label>) and interceptor group list no longer shows it — the list output is the proof, not the close command's exit code.

Report what failed or was skipped as prominently as what worked. Never report only the happy fields.

When it activates

Automatically, when a task matches. Use for browser page content, tabs, forms, SPA extraction, request overrides, page communication capture, and deployment checks. Not for native macO

About this skill

Automatic

QA & Test Engineering pack

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v8 — 11/5/2026

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

hacker-valley-media / interceptor

.agents/skills/interceptor-browser/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 11/5/2026Version v5Bundled 10 files

Illustrative Want to see what hacker-valley-media/interceptor-browser 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.

hacker-valley-media/interceptorChoose the right Interceptor surface. Use interceptor-browser for page DOM, network, browser tabs, rich editors, screenshots, and browser automation. Use interceptor-macos for native apps, browser chrome, URL bars, OS dialogs, cross-app routing, AX trees, native screenshots, Apple Events, trusted OS input, macOS Electron CDP/app web-content control, and in-process app runtime control. Use interceptor-ios for apps on an owned, unlocked, Developer-Mode iPhone. Use interceptor-research for deep mulClose match+64 pts pass-rate lift

Related skills

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

From the same source and nearby in the catalogue.

Ready to supercharge your AI agents?

SkillScore

Safety

What it changes

Track record

Use this skill

Interceptor Browser

The User Is On This Mac Too

The User Is On This Computer Too

Core Rules

Fast Path

Workflows

References

When To Switch Surfaces

Do Not Default To Troubleshooting

Completion

When it activates

SkillSafety

Source

Other skills for the same job

Related skills

More in the registry