agentleFS
rebornixrebornix/Agmente/.agents/skills/upstream-protocol-drift-watch/SKILL.md

upstream-protocol-drift-watch

Skill545 starsChanged 9 months agoMITView on GitHub

Dynamically inspect ACP and Codex upstream repos using .agmente.paths, detect protocol/API/spec drift, and produce a risk-scored regression report. Use when checking for new APIs, method changes, or protocol/spec changes after upstream updates.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 9 are distinctive, and 4 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.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If refs are not provided:

    line 23
  2. Very specific:

    Use HEAD..origin/main for each upstream repo.

    line 24
  3. Very specific:

    If origin/main is unavailable, use HEAD~1..HEAD.

    line 25
  4. Very specific:

    If HEAD~1 is unavailable, use HEAD..HEAD and report that no baseline exists.

    line 26
  5. Distinctive:

    Always fetch upstream remotes before scanning:

    line 28
  6. Distinctive:

    Read repo-local path registry:

    line 35
  7. Distinctive:

    Use user-provided refs if given.

    line 47
  8. Distinctive:

    Always perform git fetch --all --prune before resolving refs and diff range.

    line 49
  9. Very specific:

    Treat quoted/backticked values like foo/bar as method candidates.

    line 61
  10. Distinctive:

    Do not write or require a persistent baseline file unless user explicitly asks.

    line 109
  11. Distinctive:

    Do not create new scripts for this check unless user explicitly asks.

    line 110
  12. Distinctive:

    Prefer live git state and direct diff inspection each run.

    line 111
  13. Distinctive:

    Do not skip fetch unless user explicitly asks to avoid network operations.

    line 112

≈ “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

Upstream Protocol Drift Watch

Goal

Catch upstream ACP/Codex changes early, especially:

  • new RPC/API methods,
  • removed or renamed methods,
  • protocol/spec file changes that can cause Agmente regressions.

This skill is dynamic by default: always read current local repos and refs at runtime.

Inputs

Collect these from the user request when present:

  • optional from_ref and to_ref (for each repo, or shared),
  • whether to check codex, acp, or both.

If refs are not provided:

  1. Use HEAD..origin/main for each upstream repo.
  2. If origin/main is unavailable, use HEAD~1..HEAD.
  3. If HEAD~1 is unavailable, use HEAD..HEAD and report that no baseline exists.

Always fetch upstream remotes before scanning:

git -C <repo> fetch --all --prune

Required Runtime Sources

  1. Read repo-local path registry:
cat .agmente.paths
  1. Resolve:
  • AGMENTE_CODEX_REPO
  • AGMENTE_ACP_REPO
  1. Validate both paths are git repos before scanning.

Dynamic Scan Workflow

  1. Resolve diff range per repo.
  • Use user-provided refs if given.
  • Otherwise use HEAD..origin/main.
  • Always perform git fetch --all --prune before resolving refs and diff range.
  1. Collect changed files in protocol-relevant areas.
  • Codex focus paths:
    • codex-rs/app-server-protocol
    • codex-rs/app-server
  • ACP focus paths:
    • docs/schema
    • docs/protocol
    • docs/rfds
  1. Extract method-like token changes from patch.
  • Treat quoted/backticked values like foo/bar as method candidates.
  • Track added and removed candidates separately.
  1. Compare against locally supported methods.
  • Codex local method sources:
    • AppServerClient/Sources/AppServerClient/AppServerMethods.swift
    • AppServerClient/Sources/AppServerClient/AppServerEventParser.swift
  • ACP local method sources:
    • ACPClient/Sources/ACPClient/ACP/ACPMethods.swift
    • ACPClient/Sources/ACPClient/ACP/ResponseDispatcher.swift
  1. Build a risk score and severity.

Risk Scale

  • 0-19: none
  • 20-44: low
  • 45-74: medium
  • 75-100: high

Suggested scoring:

  • Path impact (max 70):
    • protocol/schema core files changed: +35 each bucket
    • app-server runtime/docs changed: +10 to +20
    • RFD/docs-only changes: +8 to +12
  • Method impact (max 30+):
    • added method not recognized locally: +12 each (cap 36)
    • removed method referenced locally: +15 each (cap 45, immediate high priority)
    • payload shape changes on core methods (thread/*, turn/*, session/*): add +10

Escalation rule:

  • Any removed locally referenced method => at least medium.
  • Multiple removed locally referenced methods => high.

Output Contract

Always return:

  1. Summary table: repo, diff range, changed file count, severity, score.
  2. Added unknown methods list.
  3. Removed supported methods list.
  4. Key changed files likely to affect compatibility.
  5. Recommended follow-up tests:
  • swift test --package-path AppServerClient
  • swift test --package-path ACPClient
  • targeted AgmenteTests protocol suites.

Guardrails

  • Do not write or require a persistent baseline file unless user explicitly asks.
  • Do not create new scripts for this check unless user explicitly asks.
  • Prefer live git state and direct diff inspection each run.
  • Do not skip fetch unless user explicitly asks to avoid network operations.

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.