agentleFS
xetenetxetenet/xete-mcp/.claude/skills/gate-retrospective/SKILL.md

gate-retrospective

Skill0 starsChanged 5 months agoMITView on GitHub

Self-improvement loop for the gate system itself. MUST be used when an open issue labeled "gate-retrospective" exists in the repo, when a gate false-positives or blocks legitimate work, when a DDR verdict later proves wrong (a shipped bug on a protected path), or monthly if none of those occur. Turns weaknesses in the gates into concrete improvement PRs against xete-agent-skills instead of silent tolerance or silent weakening.

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
  1. Very specific:

    Read the audit report (issue or fresh scripts/gate-audit.sh run), plus:

    line 20

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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

Gate Retrospective

The gates review the code. This skill reviews the gates. A gate system that can't recognize its own failure modes decays into either theater (rubber-stamps) or friction (routed around) — both are silent.

Triggers, in priority order

  1. A shipped defect on a protected path — the gate's core failure. Run immediately.
  2. An open gate-retrospective issue — the weekly audit found something. Run at session start.
  3. Friction event — a gate blocked or slowed legitimate work in the current session (false-positive pattern, redundant ceremony). Run before session end; do not just note it.
  4. Monthly heartbeat — if none of the above fired in ~30 days, run scripts/gate-audit.sh manually and retrospect on the output. Silence can mean health or mean routing-around; check which.

The loop

1. Gather evidence

Read the audit report (issue or fresh scripts/gate-audit.sh run), plus:

  • git log --oneline -- reviews/ — DDR cadence and authorship
  • The specific friction/defect event if one triggered this
  • Time cost: roughly how long did gates add to the last 3 gated merges?

2. Diagnose against the known failure modes

Failure modeEvidenceDirection of fix
TheaterNear-duplicate or thin DDRs, reflexive SHIPStrengthen review requirements OR trim protected paths — pick one, never both
FrictionFalse-positive patterns, gates on non-critical pathsNarrow patterns; move enforcement later (PR-level, not commit-level)
HolesUngated commits on main, zero gate activity during contract workWiden patterns, check branch protection toggles, check hooksPath on clones
DriftRepo layout changed, patterns didn'tRe-derive patterns from current tree

3. Doubt the fix (meta-DDR)

Changes to the gate system are themselves protected-path-grade: run doubt-driven-review on the proposed change. Key doubts to always raise:

  • Does this fix weaken the gate to relieve friction? (Weakening is sometimes right, but must be named as such, never smuggled in as "tuning.")
  • Does this fix add ceremony to relieve anxiety? (Also a failure.)
  • Will this change survive contact with an unattended agent session, or does it assume a human in the loop?

4. Ship the improvement

Open a PR against xetenet/xete-agent-skills containing: the change, the meta-DDR, and one line in the README changelog. Then re-run install.sh into affected repos (or note which repos need it in the PR body). Close or comment the triggering issue with a link.

Hard rules

  • Never silently weaken. Removing a pattern, downgrading a block to a warning, or relaxing a verdict rule requires an explicit "this weakens the gate because the cost exceeded the risk" statement in the meta-DDR, visible to the human.
  • Never silently tolerate. Grinding through a false positive without proposing a fix is a rule-6 violation (CLAUDE.md).
  • One change per retrospective. Batch fixes hide which change caused which effect. Smallest fix that addresses the diagnosed failure mode.
  • The human ratifies gate changes. Improvement PRs to this repo are proposed autonomously but merged by Chef or Jo — the gate system's own merge boundary is human.

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.