Menu
TRIGGER_DESCRIPTION. Use when USER_CONTEXT. Routes to specialists for CAPABILITIES.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Use when a user asks to read, analyze, summarize, or extract from a heavyweight file (PDF, DOCX, PPTX, XLSX, CSV, TSV). Converts the file into markdown or CSV first, generates a lightweight index, and only spends model tokens on the compressed artifact. Trigger on "read this PDF", "look through this spreadsheet", "summarize this deck", or any time raw file ingestion would waste tokens. Do NOT trigger for plain text, markdown, JSON, YAML, or source code files — those are cheap to read directly.
TRIGGER_DESCRIPTION. Use when USER_CONTEXT. Covers CAPABILITIES.
| name | SUITE_NAME |
| description | TRIGGER_DESCRIPTION. Use when USER_CONTEXT. Routes to specialists for CAPABILITIES. |
One-sentence description of what this skill suite does.
skills/specialist-a/SKILL.mdskills/specialist-b/SKILL.mdskills/specialist-a/SKILL.md| Skill | Path | Purpose | Model Tier |
|---|---|---|---|
| specialist-a | skills/specialist-a/SKILL.md | Does X | analytical |
| specialist-b | skills/specialist-b/SKILL.md | Does Y | reasoning |
Read ONLY the relevant specialist SKILL.md based on classification above. Do not pre-load multiple specialists.
Cross-skill composition: Skills may reference other installed skills by name in procedures. These references are runtime-resolved (not context-loaded) and degrade gracefully if the referenced skill is not installed. This is distinct from the cross-specialist isolation rule (§2.5), which blocks file path references between siblings.
Pass between skills as structured JSON:
{
"source_skill": "specialist-a",
"findings": [],
"confidence": "high|medium|low"
}