agentleFS
agentic-control-planeagentic-control-plane/codex-acp-plugin/skills/acp/SKILL.md

acp

Skill1 starsChanged 4 months agoMITView on GitHub

Identity, governance, and audit for all Codex tool calls via the Agentic Control Plane

  • Deletes or force-pushes

2 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Wrote their own version · 1

    1. Integralist/agent-skills★ 23same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When a user asks to set up ACP, connect to ACP, or enable governance, follow these steps:

    line 25
  2. Distinctive:

    If credentials exist, skip to Step 4 (verify).

    line 33
  3. Distinctive:

    Test the governance endpoint:

    line 59
  4. Distinctive:

    If denied, Claude sees the reason and can adapt

    line 123
  5. Distinctive:

    If the API is unreachable, the hook fails open — Claude is never blocked by ACP outages

    line 124
  6. Distinctive:

    No tool inputs are stored permanently unless the workspace has logging enabled

    line 129

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Agentic Control Plane (ACP)

ACP is the identity and governance layer for Codex. When active, every shell command is logged and governed through the user's ACP workspace before it executes. The user does not need to change how they work — ACP runs transparently via a PreToolUse hook.

Honest coverage note: Codex currently fires PreToolUse for Bash/shell commands only — file edits, web fetches, and MCP tool calls do not pass through the hook. Shell is where the blast radius lives (deploys, deletes, git, package publishing), so this still governs the actions that matter most; full-surface coverage arrives as Codex extends its hook events.

What ACP does

  • Audit logging — every tool call is recorded with identity, tool name, arguments, timestamp, and decision
  • Policy enforcement — workspace admins can allow/deny tool calls by pattern, scope, or ABAC rules
  • Rate limiting & budgets — cap tool calls per minute, per month, or by cost
  • Content scanning — detect and optionally redact PII in tool inputs
  • Team governance — multiple members per workspace, role-based access

ACP does NOT proxy or replace the user's tools. Their MCP servers, Bash commands, and file operations work exactly as before. ACP observes and governs the orchestration layer.

Setup flow — help users connect

When a user asks to set up ACP, connect to ACP, or enable governance, follow these steps:

Step 1: Check if already configured

cat ~/.acp/credentials 2>/dev/null && echo "ACP credentials found" || echo "No ACP credentials"

If credentials exist, skip to Step 4 (verify).

Step 2: Create account or log in

Open the browser to the ACP authorization page:

open "https://cloud.agenticcontrolplane.com/plugin/authorize"

Tell the user:

I've opened your browser to log in to ACP. If you don't have an account, you can sign up there with Google or email. After logging in, you'll see a token — please paste it here.

Step 3: Store credentials

Once the user provides the token, store it:

mkdir -p ~/.acp && echo "THE_TOKEN" > ~/.acp/credentials && chmod 600 ~/.acp/credentials

Confirm:

Your ACP credentials are stored securely. The governance hook is now active — every tool call in this session and future sessions will be logged to your ACP workspace.

Step 4: Verify connection

Test the governance endpoint:

curl -sf -X GET "https://api.agenticcontrolplane.com/govern/health" && echo " — ACP reachable"

Step 5: Direct to the console

Tell the user:

Your dashboard is ready. Open it to see your audit logs:

https://cloud.agenticcontrolplane.com/logs

Every tool call Claude makes — Bash commands, file edits, web fetches, MCP tools — shows up there in real-time. You can set policies, rate limits, and content scanning from the dashboard.

Console pages (for directing users)

What they wantURL
Audit logs (start here)https://cloud.agenticcontrolplane.com/logs
Policy rules (allow/deny tools)https://cloud.agenticcontrolplane.com/policies
Rate limits & budgetshttps://cloud.agenticcontrolplane.com/limits
Content/PII policyhttps://cloud.agenticcontrolplane.com/content
API keyshttps://cloud.agenticcontrolplane.com/api-keys
Team membershttps://cloud.agenticcontrolplane.com/users
Billing & usagehttps://cloud.agenticcontrolplane.com/billing

Managing ACP from within Claude

Users can ask Claude to help with ACP management. You can assist with:

Check status

TOKEN=$(cat ~/.acp/credentials 2>/dev/null)
curl -sf -H "Authorization: Bearer $TOKEN" "https://api.agenticcontrolplane.com/govern/health"

View recent tool calls

Direct the user to https://cloud.agenticcontrolplane.com/logs — the dashboard shows all tool calls with filtering by tool name, identity, time range, and decision.

Disconnect / pause governance

To temporarily disable ACP governance without uninstalling:

mv ~/.acp/credentials ~/.acp/credentials.paused

To re-enable:

mv ~/.acp/credentials.paused ~/.acp/credentials

Uninstall

rm -rf ~/.acp

Then disable the plugin: claude plugin disable agentic-control-plane

How the hook works (for user questions)

The plugin registers a PreToolUse hook that fires before every tool call:

  1. Hook reads ~/.acp/credentials for the API key
  2. Sends tool name + input to POST /govern/tool-use
  3. ACP runs a 6-layer governance pipeline (immutable rules, scope enforcement, ABAC policies, rate limits, content scanning)
  4. Returns allow or deny with a reason
  5. If denied, Claude sees the reason and can adapt
  6. If the API is unreachable, the hook fails open — Claude is never blocked by ACP outages

Important notes

  • The hook adds ~50-200ms latency per tool call (negligible for most workflows)
  • No tool inputs are stored permanently unless the workspace has logging enabled
  • ACP never modifies tool inputs or outputs — it only observes and allows/denies
  • The governance hook does NOT expose any new tools to Claude — it only monitors existing ones

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.