Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guidelines for iteratively planning tasks and changes before implementation. Use this when the user requests a non-trivial task or when you need to align on a complex implementation strategy.
.claude/skills/aiskillstore-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 390% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 149% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 14% | 0% |
Create and manage plan documents on disk. Plans stored on disk are markdown files with YAML frontmatter and free-form content. When drafting in chat, output only the plan body without frontmatter; add frontmatter only when stashing to disk. Support both implementation plans and overview/reference plans. Only write to the plans folder; do not modify the repository codebase.
$CODEX_HOME/plans or ~/.codex/plans when CODEX_HOME is not set.name and description (single-line values) for on-disk plans.# Plan.<name>.md.scripts/list_plans.py for quick summaries.scripts/read_plan_frontmatter.py to validate a specific plan.README.md, docs/, key modules) to scope requirements.scripts/create_plan.py.name and filename together.Create a plan file (body only; frontmatter is written for you). Run from the plan skill directory:
bashpython ./scripts/create_plan.py \ --name codex-rate-limit-overview \ --description "Scope and update plan for Codex rate limiting" \ --body-file /tmp/plan-body.md
Read frontmatter summary for a plan (run from the plan skill directory):
bashpython ./scripts/read_plan_frontmatter.py ~/.codex/plans/codex-rate-limit-overview.md
List plan summaries (optional filter; run from the plan skill directory):
bashpython ./scripts/list_plans.py --query "rate limit"
Use one of the structures below for the plan body. When drafting, output only the body (no frontmatter). When stashing, prepend this frontmatter:
markdown--- name: <plan-name> description: <1-line summary> ---
markdown# Plan <1-3 sentences: intent, scope, and approach.> ## Requirements - <Requirement 1> - <Requirement 2> ## Scope - In: - Out: ## Files and entry points - <File/module/entry point 1> - <File/module/entry point 2> ## Data model / API changes - <If applicable, describe schema or contract changes> ## Action items [ ] <Step 1> [ ] <Step 2> [ ] <Step 3> [ ] <Step 4> [ ] <Step 5> [ ] <Step 6> ## Testing and validation - <Tests, commands, or validation steps> ## Risks and edge cases - <Risk 1> - <Risk 2> ## Open questions - <Question 1> - <Question 2>
markdown# Plan <1-3 sentences: intent and scope of the overview.> ## Overview <Describe the system, flow, or architecture at a high level.> ## Diagrams <Include text or Mermaid diagrams if helpful.> ## Key file references - <File/module/entry point 1> - <File/module/entry point 2> ## Auth / routing / behavior notes - <Capture relevant differences (e.g., auth modes, routing paths).> ## Current status - <What is live today vs pending work, if known.> ## Action items - None (overview only). ## Testing and validation - None (overview only). ## Risks and edge cases - None (overview only). ## Open questions - None.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,439 | 3,387 | -2% | 1 | 1 | 0% | 155 | 1,514 | +877% | 0 | 0 | — |
case-02 | fail→fail | 5,300 | 6,670 | +26% | 1 | 1 | 0% | 228 | 1,642 | +620% | 0 | 0 | — |
case-12 | fail→pass | 8,798 | 2,250 | -74% | 1 | 1 | 0% | 1,366 | 1,696 | +24% | 0 | 0 | — |
case-03 | fail→fail | 3,715 | 6,114 | +65% | 1 | 1 | 0% | 513 | 1,749 | +241% | 0 | 0 | — |
case-04 | pass→pass | 7,566 | 9,829 | +30% | 1 | 1 | 0% | 1,259 | 2,328 | +85% | 0 | 0 | — |
case-05 | pass→fail | 6,587 | 26,761 | +306% | 1 | 1 | 0% | 1,107 | 4,141 | +274% | 0 | 0 | — |
case-06 | fail→pass | 3,071 | 9,068 | +195% | 1 | 1 | 0% | 482 | 2,363 | +390% | 0 | 0 | — |
case-07 | fail→pass | 8,425 | 1,479 | -82% | 1 | 1 | 0% | 1,379 | 1,534 | +11% | 0 | 0 | — |
case-08 | pass→pass | 3,656 | 1,758 | -52% | 1 | 1 | 0% | 571 | 1,592 | +179% | 0 | 0 | — |
case-09 | fail→pass | 4,666 | 2,676 | -43% | 1 | 1 | 0% | 725 | 1,802 | +149% | 0 | 0 | — |
case-10 | pass→pass | 5,236 | 2,093 | -60% | 1 | 1 | 0% | 788 | 1,648 | +109% | 0 | 0 | — |
case-11 | fail→pass | 9,308 | 2,202 | -76% | 1 | 1 | 0% | 1,496 | 1,709 | +14% | 0 | 0 | — |
case-13 | fail→pass | 12,428 | 1,878 | -85% | 1 | 1 | 0% | 1,835 | 1,597 | -13% | 0 | 0 | — |
case-14 | pass→pass | 12,747 | 3,556 | -72% | 1 | 1 | 0% | 2,000 | 1,943 | -3% | 0 | 0 | — |
case-15 | fail→pass | 11,183 | 3,243 | -71% | 1 | 1 | 0% | 1,781 | 1,937 | +9% | 0 | 0 | — |
case-16 | fail→pass | 12,331 | 2,955 | -76% | 1 | 1 | 0% | 1,616 | 1,761 | +9% | 0 | 0 | — |
case-17 | pass→pass | 8,451 | 3,801 | -55% | 1 | 1 | 0% | 1,361 | 1,939 | +42% | 0 | 0 | — |
case-18 | fail→fail | 14,399 | 7,604 | -47% | 1 | 1 | 0% | 2,464 | 1,749 | -29% | 0 | 0 | — |
case-19 | fail→pass | 14,306 | 5,803 | -59% | 1 | 1 | 0% | 1,509 | 2,193 | +45% | 0 | 0 | — |
case-20 | pass→pass | 10,653 | 4,523 | -58% | 1 | 1 | 0% | 1,545 | 1,736 | +12% | 0 | 0 | — |
case-21 | fail→pass | 12,075 | 2,647 | -78% | 1 | 1 | 0% | 1,699 | 1,687 | -1% | 0 | 0 | — |
case-22 | pass→pass | 4,448 | 1,731 | -61% | 1 | 1 | 0% | 705 | 1,597 | +127% | 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 18 counted toward the lift figure. The other 4 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 +41 percentage points is the difference between those two pass rates over the 18 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.