agentic-skill-evallisted

# Evaluating Agentic Potential Systematically evaluates every skill in a repository to determine which would produce valuable autonomous agents via the `/formula-create` → `af formula agent-gen` pipeline. ## Prerequisites - Target repository contains SKILL.md files (any structure) - AgentFactory installed (`af` CLI available) - Access to the factory's .agentfactory/AGENTS.md (current agent roster) ## Workflow ### Phase 1: Understand the Factory (do NOT skip) Read the factory's .agentfactory/AGENTS.md to understand: - What agents already exist and their types (autonomous vs interactive) - What capabilities the factory already has (design, analysis, implementation, operations) - What the dispatch pattern looks like (`af sling --agent <name> "input"`) ```bash cat "$(af root)/.agentfactory/AGENTS.md" 2>/dev/null || echo "No AGENTS.md found" ``` **Output (required):** A numbered gap list. Enumerate every capability gap the factory has. These numbers are referenced in Phase 4 — skills that cannot cite a gap number are discarded. Example: ``` 1. Security — no agent validates code safety before merge 2. Quality — no automated quality gates 3. Shipping — nothing cuts releases or changelogs ... ``` ### Phase 2: Full Skill Inventory List EVERY skill directory. No sampling. Complete coverage. ```bash find . -name "SKILL.md" -not -path "./.agentfactory/*" -not -path "./.codex/*" -not -path "./.gemini/*" | sort ``` Count total skills. This is the denominator for the evaluation