Menü
Audit recent work — trace error paths, probe edge cases, fix any bugs or gaps found
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Basierend auf der SOC-Berufsklassifikation
Performance and UX patterns discovered for ADE's Work tab, including chat/CLI/shell launch surfaces, the Work tools pane, Git/Files/iOS/App Control/Browser panels, and local-runtime-disabled perf runs. Read before editing Work tab code.
Use this skill when an agent needs to inspect or operate ADE itself through the `ade` CLI, including lanes, chats, actions, proof, runtime/socket state, or help/flag discovery.
Orchestrator-mode protocol for ADE Work-tab lead, worker, and validator chats. Use whenever the system prompt declares orchestrator-lead, orchestrator-worker, or orchestrator-validator mode. Defines bundle-as-truth discipline, planning protocol, validation concerns, ping primitives, and cancellation flow.
Autonomous PR-to-merge loop. Polls CI and review bots, fixes failures, rebases only on real conflicts, and lands the PR on main. Soft cap of 5 normal iterations plus one force-finalize iteration that bypasses review and fixes only CI. Pure loop — it does NOT run /quality or /test; run those first. Full phase logic lives in docs/playbooks/ship-lane.md.
ADE release conductor: detect whether desktop and/or iOS actually changed, bump desktop patch versions, keep iOS marketing versions fixed while bumping build numbers, ship desktop through the GitHub Actions release workflow, and distribute TestFlight builds to all beta users.
Use this skill when creating, inspecting, syncing, committing, pushing, archiving, or rebasing ADE lanes and lane worktrees through `ade lanes` and `ade git`.
| name | audit |
| description | Audit recent work — trace error paths, probe edge cases, fix any bugs or gaps found |
Audit the work you just did in this session. Don't stop at "it compiles" or "it looks right" — actively go hunting for what's wrong.
Retrace the changes. List every file you touched. For each, re-read the final state (not just the diff you remember) so you see what actually lives there now.
Trace every error path. For each changed code path: what happens on empty / nil / malformed input? When an upstream caller passes something unexpected? When a dependency throws or times out? Walk the failure branches, not just the happy path.
Hunt edge cases. Off-by-ones, empty collections, unicode, concurrency, first-run vs. repeat-run, reduce-motion / accessibility, different device/viewport sizes, streaming vs. terminal states, cancellation, partial failure. Pick the categories that actually apply to what you changed and work through them.
Check the surrounding contract. Did the change break any callers, tests, types, styling, or invariants elsewhere? Grep for references to anything you removed or renamed and confirm.
Fix what you find. For each real bug or gap, make the fix directly. For anything genuinely ambiguous, call it out rather than guessing.
Report. End with a short list: what you checked, what you fixed, and anything you deliberately left alone (and why).
Be honest — if the work was already solid, say so in one line. Don't manufacture busywork.
