Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Discover available skills in the builtin library by keyword.
.claude/skills/hezaohezao-find-skills/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 63% | 0% |
This skill helps you discover skills available in Poirot's builtin library and install user-contributed skills.
Call this skill OR the skill_search tool when:
skill_search("<keywords>")| Keyword | Likely skill | |---|---| | frontend / UI / React / Vue | frontend-design | | chart / graph / visualization | chart-visualization | | diagram / architecture | architecture-diagram | | github / PR / code review | github-code-review / github-pr-workflow | | debug / bug | systematic-debugging / python-debugpy | | test / TDD | test-driven-development | | plan / spike | plan / spike | | simplify / refactor | simplify-code |
Use this skill when the user:
Poirot ships a builtin library at:
poirot/backend/agents/skill/builtin_skills/
├── core/ # meta-skills, loaded at agent startup
├── research/ # research & intelligence skills
├── software-development/ # coding & GitHub skills
├── creative/ # visualization & design skills
└── productivity/ # docs & office skillsOnly core/ skills are auto-loaded at startup (registered as active). Skills under other categories are searchable but not auto-injected — use the skill_search tool or /skill search <query> command to find and load them on demand.
bashlist_dir("poirot/backend/agents/skill/builtin_skills/")
Use bash with grep to find skills whose name or description matches:
bashbash("grep -rl '<keyword>' poirot/backend/agents/skill/builtin_skills/ --include='SKILL.md'")
Or read a category directory and inspect each SKILL.md frontmatter:
bashlist_dir("poirot/backend/agents/skill/builtin_skills/research/")
Then read a candidate:
bashread_file("poirot/backend/agents/skill/builtin_skills/research/deep-research/SKILL.md")
Non-core skills are not auto-loaded. To use one this turn:
SKILL.md content/skill search <query> will surface + inject matched skillsautomatically once the search tool is wired)
To permanently install a user skill from an external path:
/skill install <path> [name]This copies it into skills/ (user skill storage, gitignored) and re-discovers.
When you find relevant skills, present them with:
description frontmatter)Example:
I found a skill that might help. "deep-research" (research category) guides
multi-step web research with evidence cross-verification. It's not auto-loaded
— I'll follow its guidance for this task. Want me to proceed?If no builtin skill matches:
/skill install or byauthoring skills/<name>/SKILL.md (see the skill-creator skill)
| Category | Example keywords | | -------------------- | ----------------------------------------- | | research | arxiv, deep-research, literature, osint | | software-development | debug, github, pr, codebase, subagent | | creative | chart, diagram, frontend, visualization | | productivity | docs, ppt, documentation | | core | plan, debugging, tdd, review, skill |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,848 | 8,524 | -38% | 1 | 1 | 0% | 2,269 | 1,521 | -33% | 0 | 0 | — |
case-02 | fail→fail | 14,464 | 18,224 | +26% | 1 | 1 | 0% | 2,069 | 1,524 | -26% | 0 | 0 | — |
case-03 | fail→fail | 5,146 | 9,715 | +89% | 1 | 1 | 0% | 690 | 1,607 | +133% | 0 | 0 | — |
case-04 | pass→fail | 10,778 | 6,530 | -39% | 1 | 1 | 0% | 1,991 | 1,448 | -27% | 0 | 0 | — |
case-05 | pass→fail | 9,503 | 13,553 | +43% | 1 | 1 | 0% | 1,548 | 1,437 | -7% | 0 | 0 | — |
case-06 | pass→pass | 4,489 | 5,622 | +25% | 1 | 1 | 0% | 683 | 2,035 | +198% | 0 | 0 | — |
case-07 | fail→pass | 27,643 | 7,898 | -71% | 1 | 1 | 0% | 1,424 | 2,285 | +60% | 0 | 0 | — |
case-08 | fail→pass | 11,666 | 4,474 | -62% | 1 | 1 | 0% | 1,828 | 1,803 | -1% | 0 | 0 | — |
case-09 | pass→pass | 18,445 | 3,720 | -80% | 1 | 1 | 0% | 1,581 | 1,659 | +5% | 0 | 0 | — |
case-10 | pass→pass | 17,248 | 2,691 | -84% | 1 | 1 | 0% | 2,533 | 1,465 | -42% | 0 | 0 | — |
case-11 | fail→pass | 13,910 | 8,516 | -39% | 1 | 1 | 0% | 2,236 | 2,292 | +3% | 0 | 0 | — |
case-12 | fail→pass | 16,238 | 3,527 | -78% | 1 | 1 | 0% | 1,776 | 1,732 | -2% | 0 | 0 | — |
case-13 | fail→pass | 10,057 | 13,972 | +39% | 1 | 1 | 0% | 1,391 | 2,264 | +63% | 0 | 0 | — |
case-14 | pass→fail | 26,213 | 16,882 | -36% | 1 | 1 | 0% | 2,540 | 1,490 | -41% | 0 | 0 | — |
case-15 | pass→pass | 7,270 | 7,193 | -1% | 1 | 1 | 0% | 1,113 | 2,258 | +103% | 0 | 0 | — |
case-16 | pass→pass | 10,454 | 11,199 | +7% | 1 | 1 | 0% | 1,377 | 2,841 | +106% | 0 | 0 | — |
case-17 | pass→pass | 8,601 | 6,973 | -19% | 1 | 1 | 0% | 1,170 | 1,553 | +33% | 0 | 0 | — |
case-18 | fail→pass | 17,820 | 4,442 | -75% | 1 | 1 | 0% | 928 | 1,798 | +94% | 0 | 0 | — |
case-19 | pass→pass | 11,620 | 3,803 | -67% | 1 | 1 | 0% | 1,617 | 1,774 | +10% | 0 | 0 | — |
case-20 | pass→fail | 11,662 | 7,888 | -32% | 1 | 1 | 0% | 1,680 | 1,410 | -16% | 0 | 0 | — |
case-21 | pass→pass | 9,179 | 2,643 | -71% | 1 | 1 | 0% | 1,168 | 1,500 | +28% | 0 | 0 | — |
case-22 | pass→pass | 11,213 | 10,384 | -7% | 1 | 1 | 0% | 1,300 | 2,215 | +70% | 0 | 0 | — |
case-23 | fail→pass | 8,349 | 4,317 | -48% | 1 | 1 | 0% | 1,161 | 1,808 | +56% | 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 15 counted toward the lift figure. The other 8 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 +13 percentage points is the difference between those two pass rates over the 15 comparable cases. 4 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.