Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Convert delivery findings into skill, eval, workflow, and documentation improvements.
.claude/skills/hoangnguyen0403-retro-learn/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -3% | 0% |
> !IMPORTANT] > Convert delivery findings into skill, eval, workflow, and documentation improvements.
Optional args: slug=<feature>, ticket=<id/url>, mode=interactive|autonomous|channel, channel=<id>, auto_continue=true|false, profile=business|hybrid|technical.
When the user asks to perform this workflow, execute the following steps:
Goal: Turn defects, missed expectations, and delivery friction into durable standards improvements.
session-report artifactsSKILL.md and evals/evals.json..agents/workflows.slug, root causes, skill/eval updates, follow-ups, next workflow.md# Retro: [Name] ## Evidence ## Root Causes | Finding | Category | Action | | --- | --- | --- | | [finding] | [category] | [action] | ## Skill Or Eval Updates ## Outcome Report feature_status: implemented requirement_trace: BRD-OBJ-* -> REQ-* -> AC-* -> SRS-* -> evidence completed_evidence: []; missing_evidence: []; decision_needed: []; recommended_next_workflow: none ## Next Workflow ## Follow-Ups ## Cost Report Call `get_session_cost(workflow="retro-learn")` before final handoff.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 31,002 | 14,035 | -55% | 1 | 1 | 0% | 4,794 | 3,155 | -34% | 0 | 0 | — |
case-10 | fail→pass | 10,416 | 6,213 | -40% | 1 | 1 | 0% | 1,712 | 1,499 | -12% | 0 | 0 | — |
case-02 | fail→pass | 30,552 | 16,928 | -45% | 1 | 1 | 0% | 4,618 | 3,723 | -19% | 0 | 0 | — |
case-03 | fail→pass | 25,393 | 12,317 | -51% | 1 | 1 | 0% | 4,643 | 2,820 | -39% | 0 | 0 | — |
case-04 | pass→pass | 7,421 | 3,054 | -59% | 1 | 1 | 0% | 1,248 | 1,096 | -12% | 0 | 0 | — |
case-05 | fail→pass | 10,885 | 6,329 | -42% | 1 | 1 | 0% | 1,818 | 1,764 | -3% | 0 | 0 | — |
case-15 | fail→pass | 13,757 | 2,251 | -84% | 1 | 1 | 0% | 2,216 | 953 | -57% | 0 | 0 | — |
case-06 | fail→pass | 9,223 | 4,667 | -49% | 1 | 1 | 0% | 1,541 | 1,209 | -22% | 0 | 0 | — |
case-07 | fail→pass | 7,640 | 4,897 | -36% | 1 | 1 | 0% | 1,294 | 1,414 | +9% | 0 | 0 | — |
case-08 | fail→pass | 8,065 | 7,594 | -6% | 1 | 1 | 0% | 1,344 | 1,890 | +41% | 0 | 0 | — |
case-09 | pass→pass | 14,398 | 8,367 | -42% | 1 | 1 | 0% | 2,233 | 2,036 | -9% | 0 | 0 | — |
case-11 | fail→pass | 15,334 | 4,610 | -70% | 1 | 1 | 0% | 2,535 | 1,399 | -45% | 0 | 0 | — |
case-12 | pass→pass | 8,755 | 1,594 | -82% | 1 | 1 | 0% | 1,210 | 796 | -34% | 0 | 0 | — |
case-13 | pass→fail | 9,272 | 1,853 | -80% | 1 | 1 | 0% | 1,455 | 861 | -41% | 0 | 0 | — |
case-14 | fail→pass | 7,339 | 2,602 | -65% | 1 | 1 | 0% | 1,369 | 1,085 | -21% | 0 | 0 | — |
case-16 | fail→pass | 13,099 | 4,770 | -64% | 1 | 1 | 0% | 2,258 | 1,361 | -40% | 0 | 0 | — |
case-17 | pass→fail | 12,307 | 3,347 | -73% | 1 | 1 | 0% | 2,033 | 1,157 | -43% | 0 | 0 | — |
case-18 | fail→pass | 11,821 | 2,653 | -78% | 1 | 1 | 0% | 2,014 | 972 | -52% | 0 | 0 | — |
case-19 | fail→fail | 2,211 | 4,844 | +119% | 1 | 1 | 0% | 287 | 1,337 | +366% | 0 | 0 | — |
case-20 | pass→fail | 16,943 | 2,605 | -85% | 1 | 1 | 0% | 3,943 | 877 | -78% | 0 | 0 | — |
case-21 | pass→pass | 5,860 | 5,653 | -4% | 1 | 1 | 0% | 930 | 1,417 | +52% | 0 | 0 | — |
case-22 | fail→pass | 8,091 | 2,616 | -68% | 1 | 1 | 0% | 1,239 | 1,016 | -18% | 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 +50 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 cases got worse with the skill loaded, and they are 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.