LoopEngineering-CrashCourse

Reviews pending PRs, excludes stale/unfinished work, combines the rest into a release batch built from main. Read-only — reports the batch, never actually releases.

Writes a scoped spec packet for one ticket, reviewed by a genuinely independent adversarial pass that also regression-checks prior findings. Read-only outside the draft — never writes the canonical spec.

Defines every required outcome and its evidence before acting on long-running work, then audits each requirement as proved/weak/missing/ contradicted. Only run when explicitly asked for a completion audit.

General-purpose harness for scheduled repo work — stage in an isolated worktree, verify with a genuinely separate second session, ship only after a pass. Read-only — reports, never delivers.

Runs an adversarial review cycle against a PR (Claude implements/fixes, Codex or a subagent stand-in reviews), up to 5 iterations. Read-only on the real PR — drafts fixes, never pushes them.

Turns a ticket or bug report into a review-ready patch. Read-only — reproduces, roots out the cause, drafts a fix, reports it. Never opens a PR, never folds in unrelated refactors.

Inspects branches, PRs, commits, and worktrees for what's stale vs. valuable. Read-only. Never deletes uncertain work or someone else's PR without confirmation.

Reviews production logs for actionable errors. Read-only — drafts a traced, verified fix as a report, never opens a PR. Never copies credentials or PII out of the logs.

Optimizes page-load speed until every route is under 50ms. Read-only — drafts candidate optimizations, never applies them.

Scans dependencies for CVEs, ranks by severity/reachability, drafts the smallest credible patch. Read-only — never applies a patch or accepts risk on its own.

Finds flaky tests, roots out the real cause, drafts a fix — never a blind sleep or retry. Read-only — reports, doesn't commit, at L1.

Adds tests until the suite reaches 100% coverage. Drafts candidate tests in a throwaway worktree and reports them — never commits a test file itself.

Reviews the codebase against its docs. Read-only. Flags stale documentation where the code has moved on — never edits a doc or opens a PR at L1.

Daily or per-tag changelog drafting. Read-only on CHANGELOG.md — drafts entries for every merged PR since the last beat, never commits them.

Periodic issue triage. Read-only. Produces the 5-line report, issues only (no PRs or CI).

Sweeps for loose ends after merges. Read-only. Flags merged branches, issues a merge should have closed, and orphaned preview environments — never deletes or closes anything itself.

Nightly sweep of outdated dependencies. Tests each candidate update in a throwaway worktree and reports the result. Read-only outside its own sandbox — does not open a PR until promoted past L1.

Watches main's CI. On a red run, classifies flaky vs real and drafts a suggested fix in a report. Read-only — does not touch a worktree or open a PR until promoted past L1.

Polls every open PR on a fast schedule. Read-only. Flags merge conflicts, red CI, and stale reviews — never comments or merges.

Morning repo triage. Read-only. Produces the 5-line report.

<One line naming WHEN this skill applies, in the words a task would use. e.g. "Morning repo triage. Read-only. Produces the 5-line report.">

Check token budget and run-log spend before and after a loop run. Enforces early exit when over budget or when there is no actionable work.

Read loop-constraints.md at the start of every run and enforce every rule. This skill runs BEFORE triage or any action skill. Constraints are binding.