agentleFS
Sign inStart free
MaximSTHMaximSTH/meta-layer-starter/.agents/skills/refresh-vendor/SKILL.md

refresh-vendor

Skill2 starsChanged 5 months agoMITView on GitHub

Walk the vendor-knowledge refresh protocol on demand OR on the weekly calendar reminder OR on drift-on-encounter — fetch sources, surface drift, human gate. Change-marker semantics: no-op walks write the ledger, no PR.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Apply — conditional on drift: If drift was detected and approved: edit <vendor>.md.

    line 19

≈ “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

/refresh-vendor

Substance lives in markdowns/protocols/refresh-vendor.md.

Steps:

  1. Discriminate kickoff vs iteration per auto-trigger-discriminator.md. If iteration / polish / investigation, exit silently — conversation continues without the protocol-walk. Explicit /refresh-vendor invocation bypasses the discriminator entirely (walks the full protocol). The other three entry points from refresh-vendor.md — weekly calendar reminder, drift-on-encounter, rate-limit / flag-mismatch failure in scripts/cross-vendor-review.sh — are kickoff by definition (no discriminator ambiguity to resolve).
  2. Pick the vendor. Resolve to markdowns/agents/vendor-knowledge/<vendor>.md (claude-code, codex-cli, antigravity-cli). Unknown name = stop and ask.
  3. Read the knowledge file. Note volatility tags ([STABLE] / [MEDIUM] / [VOLATILE]), source URLs in §-level citations, and the last-verified frontmatter date.
  4. Fetch sources. Pull every cited URL via WebFetch. Read official changelogs / release notes / docs anchored in the file. Don't invent capabilities the vendor doesn't claim.
  5. Produce a drift report (drift-detected only) per the protocol's "The walk" step 4: added capabilities, removed capabilities, changed behavior (renamed flags, deprecations, breaking changes), volatility re-tagging. One bullet per claim with the source channel cited. A walk that finds nothing skips the report and goes to step 8.
  6. Walk FCPSS coverage (drift-detected only) per fcpss-gate.md. Work shape = research. Output = updated knowledge file + drift memo (drift-detected) OR ledger entry only (no-op). No-op walks skip this step. Concrete artifacts only — paste-fill "N/A" defeats the gate per the anchored-observations rule in cross-vendor-review.md.
  7. Human gate. Surface the drift report in chat (drift-detected walks only). Supervisor approves / declines / amends each delta. NO writes to the knowledge file before approval.
  8. Apply — conditional on drift. If drift was detected and approved: edit <vendor>.md; bump last-verified; re-run dependent scripts. If no drift was found: do NOT edit <vendor>.md; append one line to markdowns/agents/refresh-log.md per the protocol's "The walk" step 7; the walk ends here (no PR).
  9. Ship per tier (drift-detected only). Tier 2 typical — supervision pattern via supervision.md: worker self-check, cross-vendor review by a peer vendor, chat-checkpoint, wait for "ship it". No-op walks have nothing to ship.

Auto-trigger fires on (a) weekly calendar reminder (you open Claude Code and type /refresh-vendor <vendor>), (b) user invocation in chat (/refresh-vendor <vendor> typed directly), (c) drift-on-encounter (per-volatility ceiling exceeded since last-verified AND a vendor capability is referenced in-session), or (d) rate-limit / flag-mismatch failure in scripts/cross-vendor-review.sh traceable to a stale flag table — mandatory refresh per the failure-modes table in cross-vendor-review.md. Change-marker semantics: walks that find nothing append to the ledger and exit without editing the knowledge file.

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