agentleFS
TheVellerTheVeller/claude-skills/skills/agent-operations/govern-agentic-os/SKILL.md

govern-agentic-os

Skill4 starsChanged 13 days agoMITView on GitHub

Audit, plan, execute, check, and hand off governance work for this Agentic OS/Claudesidian repository. Use when reconciling the repository with Linear Documents, auditing documentation or architecture, validating phase structure or historical-issue absorption, fixing cross-session drift, deciding where documentation belongs, or preparing an Agentic OS handoff.

  • Reads credentials

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 9 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

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

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

    If the user does not select a mode, infer it from the request.

    line 20
  2. Distinctive:

    Read governance-contract.md for every mode.

    line 24
  3. Very specific:

    Read audit-protocol.md for audit, plan, or repo-wide execute work.

    line 25
  4. Very specific:

    Read .config/agentic-os-governance.json when it exists.

    line 26
  5. Distinctive:

    Read the relevant Linear Document or issue before judging its local projection.

    line 27
  6. Very specific:

    Use memory/Graphify/OpenWiki only to discover pointers.

    line 28
  7. Distinctive:

    Read source files semantically and generated/vendor artifacts mechanically.

    line 35
  8. Very specific:

    Never read 01_Journal/, 03_Areas/, 04_Resources/, secret-like paths, credentials, tokens, or local environment files during this workflow.

    line 40
  9. Distinctive:

    Start from audit evidence, not memory summaries.

    line 44
  10. Distinctive:

    Keep Linear Documents canonical, issues as change records, and local mirrors generated or explicitly code-adjacent.

    line 55
  11. Very specific:

    Run proportional tests and check before closing work.

    line 56
  12. Distinctive:

    Run the single repository checker:

    line 61
  13. Distinctive:

    Put the handoff in the operating system's temporary directory, not the repository.

    line 77
  14. Distinctive:

    Update Linear only with explicit authorization.

    line 81
  15. Very specific:

    After any change to .claude/commands/, .claude/skills/, or .agents/skills/, follow the repository's ADK sync instructions.

    line 89
  16. Distinctive:

    Write current project narrative in Spanish.

    line 93

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Govern Agentic OS

Govern the repository without creating a second source of truth. Treat Linear Documents as durable project canon, repository files as executable or offline artifacts, and memory systems as pointer/derived-index layers.

Select one mode

ModeDefault authorityPurpose
auditRead-onlyCollect evidence and classify drift.
planRead-onlyPropose a gated remediation plan.
executeMutating, explicitly authorizedApply only an approved plan.
checkRead-onlyRun deterministic coherence checks.
handoffRead-only by defaultCompact state into links, evidence, and next actions.

If the user does not select a mode, infer it from the request. Never infer authorization for execute from an audit, diagnosis, review, or plan request.

Load only the needed context

  1. Read governance-contract.md for every mode.
  2. Read audit-protocol.md for audit, plan, or repo-wide execute work.
  3. Read .config/agentic-os-governance.json when it exists. Treat it as the machine-readable map of Linear IDs, permitted mirrors, exclusions, and phase invariants; do not hard-code those IDs elsewhere.
  4. Read the relevant Linear Document or issue before judging its local projection.
  5. Use memory/Graphify/OpenWiki only to discover pointers. Verify every material claim at its source.

Apply the safe workflow

Audit

  1. State the scope and exclusions before reading files.
  2. Read source files semantically and generated/vendor artifacts mechanically.
  3. Inspect nested repositories for architecture, documentation, integration, and repository health; deep-review code only when evidence reveals drift or risk.
  4. Record each finding with severity, source, observed state, expected authority, evidence, and proposed disposition.
  5. Correct nothing. Publish findings to Linear only when the user explicitly authorizes the write.

Never read 01_Journal/, 03_Areas/, 04_Resources/, secret-like paths, credentials, tokens, or local environment files during this workflow.

Plan

  1. Start from audit evidence, not memory summaries.
  2. Preserve the authority split in the governance contract.
  3. Group work into small phases with one verifiable gate per phase.
  4. Correct P0/P1 findings in execution; route P2/P3 findings to prioritized Linear backlog.
  5. Show exact Linear and filesystem mutation targets. Do not apply the plan.

Execute

  1. Require an approved plan and resolve its exact targets.
  2. Re-read each target immediately before mutation.
  3. Apply the smallest coherent change; never auto-pull or auto-fix unrelated drift.
  4. Keep Linear Documents canonical, issues as change records, and local mirrors generated or explicitly code-adjacent.
  5. Run proportional tests and check before closing work.
  6. Consolidate factual evidence in the relevant issue body only after verification. Do not create a parallel implementation log in comments.

Check

Run the single repository checker:

npm run agentic-os:check
npm run agentic-os:check -- --live
npm run agentic-os:check -- --json

Local mode is the offline default. It validates the governance config, generated-projection checksums, phase metadata, package/script safety, intention invariants, and ADK mirrors without reading secrets or making network calls. --live adds read-only Linear Document reachability and requires LINEAR_API_KEY; runtime tooling remains an explicit npm run tooling:check because it crosses into user-home runtime configuration. Neither mode writes or pulls. Interpret exit codes as:

  • 0: checks passed; warnings may remain.
  • 1: coherence drift.
  • 2: invalid invocation or missing live precondition.

Handoff

  1. Put the handoff in the operating system's temporary directory, not the repository.
  2. Reference Linear Documents, issues, commits, diffs, and files instead of copying them.
  3. Include scope, verified state, remaining gates, exact next command, and suggested skills.
  4. Redact secrets and personal information.
  5. Update Linear only with explicit authorization.

Enforce the session contract

At session start, run local check and report drift without correcting it. Open the applicable Linear Document before planning or editing.

At session close, run local and live checks when authentication is available, update the relevant Linear issue body with implementation evidence, then prepare a pointer-based handoff. Comments are for exceptional human discussion, not canonical status or evidence. Never mark an issue Done while a required live check fails.

After any change to .claude/commands/, .claude/skills/, or .agents/skills/, follow the repository's ADK sync instructions. Do not run sync merely by invoking this skill.

Keep outputs English-ready

Write current project narrative in Spanish. Use English for filenames, command names, JSON keys, stable identifiers, and contract headings. Do not maintain parallel translations. Defer full English migration, privacy review, link review, and licensing to the explicit open-source publication gate.

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.