agentleFS
Sign inStart free
ARCASSystemsARCASSystems/FounderOS/skills/approval-gates/SKILL.md

approval-gates

Skill7 starsChanged 19 days agoMITView on GitHub

Check whether an action needs approval before doing it. Say "do I need approval for this" or "should I ask first". Triggers when you are about to send, publish, pay, sign, delete, or push something public. Reads `rules/approval-gates.md` and classifies the action as auto-runnable, ask-first, or refused. Other skills call this internally.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 6 are distinctive, and 3 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.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read rules/approval-gates.md.

    line 16
  2. Distinctive:

    If neither file exists, default to ask-first for any external, financial, public, destructive, or data-sensitive action.

    line 18
  3. Very specific:

    Read the relevant gate from rules/approval-gates.md.

    line 39
  4. Very specific:

    If the gate is ambiguous, choose ASK FIRST.

    line 44
  5. Distinctive:

    If an action changes state after approval, log the outcome to the right file:

    line 61
  6. Distinctive:

    If the user rejects the action, log only if the decision affects future behavior.

    line 67
  7. Distinctive:

    Do not auto-approve because a similar action happened before.

    line 71
  8. Distinctive:

    Do not ask for approval on actions clearly below threshold.

    line 72
  9. Distinctive:

    Do not proceed silently on partial approval.

    line 73

≈ “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

Approval Gates

Runs on: reasoning - reads your files and reasons; any capable agent can run this.

Pre-Read

Read rules/approval-gates.md. If the installed OS has no customized copy, read templates/rules/approval-gates.md as the fallback.

If neither file exists, default to ask-first for any external, financial, public, destructive, or data-sensitive action.

Action Classes

Classify the requested action into one of these groups:

ClassDefault Gate
Internal note, log entry, or draftAuto-runnable
Read-only scan or local analysisAuto-runnable
Edit to operating files with user dataAsk first
External send, publish, payment, contract, invoice, or proposalAsk first
Public repo push, public release, or public package updateAsk first
Delete, hard reset, force push, secret exposure, or AI attribution in commit historyRefuse unless the user explicitly owns the risk and the action is allowed by repo rules

The local rules/approval-gates.md wins over this default table.

Procedure

  1. State the action in one sentence.
  2. Classify the action type.
  3. Read the relevant gate from rules/approval-gates.md. If that file is absent, fall back to templates/rules/approval-gates.md. If neither exists, use the default table above.
  4. Return one of three calls:
    • AUTO-RUN - proceed and log if the action changes state.
    • ASK FIRST - show the exact action and wait for yes.
    • REFUSE - explain the blocked action and the safer path.
  5. If the gate is ambiguous, choose ASK FIRST.

Approval Artifact

For ask-first actions, present:

Approval needed: <action>
Gate: <rule name or section>
Why approval is needed: <one sentence>
What will change: <files, external party, public surface, money, or data>
Rollback path: <how to undo, or none>
Question: Proceed? yes / no

Logging

If an action changes state after approval, log the outcome to the right file:

  • brain/log.md for operating actions.
  • context/decisions.md for resolved decisions.
  • context/clients.md for revenue or client state.

If the user rejects the action, log only if the decision affects future behavior.

Rules

  • Do not auto-approve because a similar action happened before.
  • Do not ask for approval on actions clearly below threshold.
  • Do not proceed silently on partial approval.

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.

Check your setup, free