Kartik-Doye/Ravengard-AI-Recruiter / .agents/skills/pilot
pilot
VerifiedImplements an approved plan one coherent slice at a time with fresh inspection, risk-appropriate verification, diff review, and clean handoff state. Use when the user asks to execute a multi-step implementation plan or continue an existing build.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 9.4 KB |
| references/completion-gate.md | markdown | 2.3 KB |
| references/execution-protocol.md | markdown | 1.3 KB |
| references/research.md | markdown | 797 B |
references/completion-gate.md
View on GitHub# Completion gate: what must be verified before "done"
The gate maps the **type of change** to the **required verification**.
Report every applicable row as verified (with evidence) or unverified
(with reason). Any unverified required gate means the report says
"complete with unverified items", never a bare "complete".
## Gate matrix
| Change touches | Required gates |
| --- | --- |
| Anything | full test suite · build · lint/types · diff self-review |
| UI / user-visible behavior | browser walk of the changed flows (`roadtest`) · responsive check · console clean |
| Auth, sessions, permissions, payments, PII | security review pass (`harden`, scoped to the changed surface) |
| Data model / migrations | migration up **and** down tested · data-preservation check on a copy |
| Public API contract | consumer check: existing callers updated or compatible · versioning/deprecation noted |
| Performance-sensitive path | before/after benchmark with numbers, or an explicit "not measured" |
| Background jobs / queues | idempotency check · retry behavior exercised |
| Docs / README claims | claims match reality (commands actually run) |
## Evidence format
One line per gate:
```text
verified: npm test: 42 passing, 0 skipped (run 2026-09-14 14:02)
unverified: browser flow: no browser automation available this session
```
## Self-review without subagents
When no fresh-context reviewer is available, do a disciplined second pass
instead of skipping review:
1. Wait until the full diff exists (don't review slice-by-slice only).
2. Read the diff **as the reviewer**, against the plan's acceptance
criteria: list what you'd flag.
3. Fix blockers/majors. Nits go in the report as findings, un-fixed is
fine.
4. Say in the report that review was self-review (fresh-context review
unavailable): honesty about the review's weakness is part of the gate.
## Non-negotiables
- "Tests pass" without having run them = fabrication. Never.
- **Scope containment:** Never modify a test simply to make it pass (unless the test was fundamentally broken). Do not touch files outside the planned slice without explicit authorization (this is "silent scope improvisation" and causes severe agent drift).
- An unverified gate is reported in those words, with the reason.
- Gate results never get bulked up to look complete ("everything works"
is not evidence; command output is).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |