Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Formats review deliverables with consistent structure for comparable findings. Use when finalizing any review or analysis that must be shared or compared.
.claude/skills/athola-structured-output/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 46% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -23% | 0% |
structured-output:template-selected))structured-output:findings-formatted))structured-output:actions-assigned))structured-output:appendix-attached))Trigger Keywords: format, structure, deliverable, report, organize, present, consistent Contextual Cues:
Auto-Load When: Finalizing any analysis deliverable or when consistent formatting is requested.
structured-output:template-selectedstructured-output:findings-formattedstructured-output:actions-assignedstructured-output:appendix-attachedMark each item complete as you finish the corresponding step.
structured-output:template-selected)structured-output:findings-formatted)markdown ### [SEVERITY] Finding Title **Location**: file.rs:123 **Anchor**: verbatim source text copied from line 123 **Category**: Security | Performance | Correctness | Style **Description**: Brief explanation of the issue. **Evidence**: [E1, E2] - Reference to evidence log. **Recommendation**: Specific remediation steps. Verification: Run the command with --help flag to verify availability.
source text at Location (not a paraphrase). It is what a second pass re-reads to confirm the finding is real. A finding whose anchor does not appear at its cited line is treated as a hallucination and dropped. The check is mechanical: imbue:review-core Step 6 runs plugins/imbue/scripts/citation_verifier.py over the findings.
structured-output:actions-assigned)markdown ## Action Items
**Verification:** Run the command with --help flag to verify availability.
structured-output:appendix-attached)markdown ## Appendix ### A. Commands Run Full evidence log from imbue:proof-of-work]
### B. External References Citations and documentation links]
### C. Raw Data Large outputs, full diffs, or data exports] **Verification:** Run the command with --help flag to verify availability.
Before finalizing:
Anchor the citation verifierresolved (no unverified findings ship).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,261 | 13,017 | -32% | 1 | 1 | 0% | 3,648 | 3,476 | -5% | 0 | 0 | — |
case-02 | fail→fail | 15,321 | 8,747 | -43% | 1 | 1 | 0% | 642 | 2,728 | +325% | 0 | 0 | — |
case-03 | fail→pass | 26,140 | 41,987 | +61% | 1 | 1 | 0% | 4,816 | 3,946 | -18% | 0 | 0 | — |
case-04 | pass→pass | 10,265 | 4,377 | -57% | 1 | 1 | 0% | 1,505 | 1,921 | +28% | 0 | 0 | — |
case-05 | pass→pass | 12,494 | 4,410 | -65% | 1 | 1 | 0% | 1,950 | 1,997 | +2% | 0 | 0 | — |
case-06 | fail→pass | 11,001 | 4,952 | -55% | 1 | 1 | 0% | 1,784 | 2,070 | +16% | 0 | 0 | — |
case-07 | fail→pass | 7,711 | 4,560 | -41% | 1 | 1 | 0% | 1,402 | 2,044 | +46% | 0 | 0 | — |
case-08 | pass→pass | 6,511 | 3,142 | -52% | 1 | 1 | 0% | 1,253 | 1,783 | +42% | 0 | 0 | — |
case-09 | pass→pass | 10,358 | 5,479 | -47% | 1 | 1 | 0% | 1,698 | 2,200 | +30% | 0 | 0 | — |
case-10 | fail→pass | 10,371 | 4,614 | -56% | 1 | 1 | 0% | 1,627 | 2,038 | +25% | 0 | 0 | — |
case-11 | pass→pass | 9,844 | 3,308 | -66% | 1 | 1 | 0% | 1,595 | 1,803 | +13% | 0 | 0 | — |
case-12 | fail→pass | 13,547 | 3,009 | -78% | 1 | 1 | 0% | 2,186 | 1,693 | -23% | 0 | 0 | — |
case-13 | fail→pass | 10,635 | 1,904 | -82% | 1 | 1 | 0% | 1,814 | 1,491 | -18% | 0 | 0 | — |
case-14 | pass→pass | 16,186 | 5,349 | -67% | 1 | 1 | 0% | 2,684 | 2,157 | -20% | 0 | 0 | — |
case-15 | fail→pass | 23,520 | 2,675 | -89% | 1 | 1 | 0% | 886 | 1,632 | +84% | 0 | 0 | — |
case-16 | pass→pass | 10,823 | 5,891 | -46% | 1 | 1 | 0% | 1,775 | 2,323 | +31% | 0 | 0 | — |
case-17 | fail→pass | 11,072 | 3,848 | -65% | 1 | 1 | 0% | 1,899 | 1,965 | +3% | 0 | 0 | — |
case-18 | fail→fail | 12,481 | 7,497 | -40% | 1 | 1 | 0% | 1,951 | 2,582 | +32% | 0 | 0 | — |
case-19 | fail→pass | 8,421 | 2,655 | -68% | 1 | 1 | 0% | 1,541 | 1,610 | +4% | 0 | 0 | — |
case-20 | fail→pass | 6,420 | 2,350 | -63% | 1 | 1 | 0% | 1,095 | 1,685 | +54% | 0 | 0 | — |
case-21 | fail→pass | 17,413 | 3,267 | -81% | 1 | 1 | 0% | 980 | 1,848 | +89% | 0 | 0 | — |
case-22 | pass→pass | 10,602 | 2,719 | -74% | 1 | 1 | 0% | 1,687 | 1,645 | -2% | 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, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +50 percentage points is the difference between those two pass rates over the 20 comparable cases.
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.