Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Analyze codebase with parallel mapper agents to produce .planning/codebase/ documents
.claude/skills/coco-research-gsd-map-codebase/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -57% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -15% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -34% | 0% |
<objective> Analyze existing codebase using parallel gsd-codebase-mapper agents to produce structured codebase documents.
Each mapper agent explores a focus area and writes documents directly to .planning/codebase/. The orchestrator only receives confirmations, keeping context usage minimal.
Output: .planning/codebase/ folder with 7 structured documents about the codebase state. </objective>
<execution_context> @$HOME/.claude/get-shit-done/workflows/map-codebase.md </execution_context>
<context> Focus area: $ARGUMENTS (optional - if provided, tells agents to focus on specific subsystem)
Load project state if exists: Check for .planning/STATE.md - loads context if project already initialized
This command can run:
</context>
<when_to_use> Use map-codebase for:
Skip map-codebase for:
</when_to_use>
<process>
</process>
<success_criteria>
</success_criteria>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | pass→pass | 12,060 | 11,253 | -7% | 1 | 1 | 0% | 2,080 | 1,685 | -19% | 0 | 0 | — |
case-01 | fail→fail | 16,794 | 8,441 | -50% | 1 | 1 | 0% | 1,967 | 817 | -58% | 0 | 0 | — |
case-02 | fail→fail | 5,235 | 40,747 | +678% | 1 | 1 | 0% | 360 | 9,018 | +2405% | 0 | 0 | — |
case-03 | fail→fail | 8,988 | 9,880 | +10% | 1 | 1 | 0% | 142 | 975 | +587% | 0 | 0 | — |
case-04 | fail→fail | 13,038 | 9,615 | -26% | 1 | 1 | 0% | 209 | 801 | +283% | 0 | 0 | — |
case-05 | fail→fail | 26,784 | 16,235 | -39% | 1 | 1 | 0% | 5,663 | 1,155 | -80% | 0 | 0 | — |
case-06 | pass→fail | 11,068 | 19,986 | +81% | 1 | 1 | 0% | 817 | 1,049 | +28% | 0 | 0 | — |
case-07 | fail→pass | 14,356 | 2,959 | -79% | 1 | 1 | 0% | 2,448 | 1,054 | -57% | 0 | 0 | — |
case-08 | fail→fail | 11,323 | 1,953 | -83% | 1 | 1 | 0% | 1,808 | 900 | -50% | 0 | 0 | — |
case-09 | fail→pass | 16,106 | 7,792 | -52% | 1 | 1 | 0% | 1,843 | 1,065 | -42% | 0 | 0 | — |
case-10 | pass→pass | 11,112 | 12,436 | +12% | 1 | 1 | 0% | 859 | 908 | +6% | 0 | 0 | — |
case-11 | pass→pass | 20,669 | 8,807 | -57% | 1 | 1 | 0% | 2,785 | 2,167 | -22% | 0 | 0 | — |
case-12 | fail→pass | 15,665 | 3,212 | -79% | 1 | 1 | 0% | 1,615 | 1,230 | -24% | 0 | 0 | — |
case-13 | pass→fail | 4,109 | 8,871 | +116% | 1 | 1 | 0% | 649 | 989 | +52% | 0 | 0 | — |
case-14 | fail→pass | 18,680 | 7,646 | -59% | 1 | 1 | 0% | 1,985 | 1,692 | -15% | 0 | 0 | — |
case-16 | pass→pass | 10,932 | 10,142 | -7% | 1 | 1 | 0% | 960 | 1,280 | +33% | 0 | 0 | — |
case-17 | pass→pass | 20,197 | 11,878 | -41% | 1 | 1 | 0% | 2,407 | 1,794 | -25% | 0 | 0 | — |
case-18 | fail→pass | 22,583 | 11,412 | -49% | 1 | 1 | 0% | 2,416 | 1,605 | -34% | 0 | 0 | — |
case-19 | fail→pass | 7,167 | 8,008 | +12% | 1 | 1 | 0% | 1,141 | 1,088 | -5% | 0 | 0 | — |
case-20 | fail→pass | 19,338 | 11,454 | -41% | 1 | 1 | 0% | 2,334 | 1,707 | -27% | 0 | 0 | — |
case-21 | pass→pass | 16,151 | 10,353 | -36% | 1 | 1 | 0% | 1,779 | 1,403 | -21% | 0 | 0 | — |
case-22 | fail→pass | 13,194 | 8,759 | -34% | 1 | 1 | 0% | 1,557 | 1,189 | -24% | 0 | 0 | — |
case-23 | pass→pass | 15,524 | 11,309 | -27% | 1 | 1 | 0% | 1,395 | 1,362 | -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. 23 cases were attempted, and 17 counted toward the lift figure. The other 6 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 +26 percentage points is the difference between those two pass rates over the 17 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.