Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Pin Pilot executor failures against known v2.149.x patterns before exploring adjacent systems. Invoke when user says "pilot failed", "executor stuck", "epic stuck", "autopilot didn't fire", or similar pilot-failure phrasing.
.claude/skills/qf-studio-pilot-debug/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 24% | 0% |
When invoked, state the hypothesis FIRST in this exact order before running any investigation tool. Each pattern was patched in v2.149.1–.3 and has high prior probability of recurring.
startup gate log lines
/tmp/pilot-worktree-*and whether git rev-parse --git-common-dir resolves correctly
executor/worktree.go:86 vs cmd/pilot/repo_allowlist.go:53allChildrenDone(...) invocation site for emptychild-set handling
epic.go:923CommitSHAor PRUrl are being flagged as failed
cmd/pilot/handlers.go, cmd/pilot/commands.goState one of:
match: <pattern N> — proceed with the per-pattern diagnostic above.novel: <one-line reason none match> — write a failing regression testFIRST, then escalate.
Do NOT touch session dirs, daemon config, queue state, or DB until all 4 known patterns are ruled out. Past sessions burned cycles wandering through session_dirs/ and daemon config before the actual bug surfaced.
.agent/knowledge/memories/pitfalls/mem-pilot-001.md — silent gate.agent/knowledge/memories/pitfalls/mem-pilot-002.md — worktree path.agent/knowledge/memories/pitfalls/mem-pilot-003.md — vacuous truth.agent/knowledge/memories/pitfalls/mem-pilot-004.md — IsEpic guard| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,601 | 4,364 | -68% | 1 | 1 | 0% | 2,106 | 1,257 | -40% | 0 | 0 | — |
case-02 | fail→fail | 13,566 | 7,426 | -45% | 1 | 1 | 0% | 2,008 | 1,888 | -6% | 0 | 0 | — |
case-03 | fail→fail | 12,797 | 6,604 | -48% | 1 | 1 | 0% | 2,012 | 1,835 | -9% | 0 | 0 | — |
case-04 | pass→pass | 4,387 | 4,745 | +8% | 1 | 1 | 0% | 784 | 1,424 | +82% | 0 | 0 | — |
case-05 | pass→pass | 9,465 | 3,682 | -61% | 1 | 1 | 0% | 1,517 | 1,206 | -21% | 0 | 0 | — |
case-06 | fail→pass | 9,799 | 1,612 | -84% | 1 | 1 | 0% | 1,382 | 793 | -43% | 0 | 0 | — |
case-07 | fail→pass | 14,339 | 4,034 | -72% | 1 | 1 | 0% | 1,944 | 1,322 | -32% | 0 | 0 | — |
case-08 | fail→pass | 12,158 | 8,783 | -28% | 1 | 1 | 0% | 1,772 | 2,254 | +27% | 0 | 0 | — |
case-09 | fail→pass | 8,965 | 9,154 | +2% | 1 | 1 | 0% | 1,389 | 2,289 | +65% | 0 | 0 | — |
case-10 | fail→pass | 9,440 | 6,121 | -35% | 1 | 1 | 0% | 1,352 | 1,678 | +24% | 0 | 0 | — |
case-11 | fail→pass | 8,020 | 6,767 | -16% | 1 | 1 | 0% | 1,179 | 1,787 | +52% | 0 | 0 | — |
case-12 | fail→pass | 14,880 | 5,204 | -65% | 1 | 1 | 0% | 2,346 | 1,521 | -35% | 0 | 0 | — |
case-13 | fail→fail | 9,134 | 6,342 | -31% | 1 | 1 | 0% | 1,333 | 1,760 | +32% | 0 | 0 | — |
case-14 | fail→pass | 10,031 | 4,434 | -56% | 1 | 1 | 0% | 1,512 | 1,253 | -17% | 0 | 0 | — |
case-15 | fail→fail | 15,978 | 3,901 | -76% | 1 | 1 | 0% | 2,346 | 1,152 | -51% | 0 | 0 | — |
case-16 | fail→fail | 12,501 | 5,116 | -59% | 1 | 1 | 0% | 1,896 | 1,409 | -26% | 0 | 0 | — |
case-17 | pass→fail | 12,175 | 4,723 | -61% | 1 | 1 | 0% | 1,939 | 1,352 | -30% | 0 | 0 | — |
case-18 | fail→pass | 4,979 | 3,747 | -25% | 1 | 1 | 0% | 703 | 1,168 | +66% | 0 | 0 | — |
case-19 | fail→fail | 13,805 | 5,619 | -59% | 1 | 1 | 0% | 2,306 | 1,725 | -25% | 0 | 0 | — |
case-20 | fail→pass | 8,715 | 8,028 | -8% | 1 | 1 | 0% | 1,358 | 2,385 | +76% | 0 | 0 | — |
case-21 | fail→pass | 6,451 | 6,974 | +8% | 1 | 1 | 0% | 1,099 | 1,964 | +79% | 0 | 0 | — |
case-22 | pass→pass | 6,791 | 8,130 | +20% | 1 | 1 | 0% | 945 | 2,170 | +130% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +45 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.