prism

Query the skill registry to find relevant architectural skills matching what the user wants to build. Searches across configured registries with graceful fallback to local sources.

Query the skill registry to find relevant architectural skills matching the user codebase, then audit the code against matched skill recommendations.

Run a deep architectural analysis of an agentic AI codebase using a phased approach — index, parallel clustered analysis, and synthesis.

Extract reusable, framework-agnostic skills from an agent codebase analysis report and generate registry-ready skill files. TRIGGER when: a full_report.md has been generated by `/analyze-agent-codebase` (Claude Code) or `@analyze-agent-codebase` (Cursor), the user runs `/extract-skills` (Claude Code) or `@extract-skills` (Cursor), the user asks to extract skills from an analysis.

Extract design decisions, trade-offs, and non-obvious behaviors from a codebase's current source code. TRIGGER when: starting the design analysis pipeline, running `/mine-design` (Claude Code) or `@mine-design` (Cursor) before `/synthesize-decisions` (Claude Code) or `@synthesize-decisions` (Cursor), capturing the reasoning behind current architectural choices, analyzing a codebase snapshot for patterns that would surprise a developer arriving cold.

Run the codebase analysis pipeline — routes to agentic or general path, one step at a time, clearing context between steps (`/clear` in Claude Code, or clear context in Cursor) to prevent context blowup.

Promote design decisions from _analysis/design.md into normative practice skills — guidance that loads before a developer makes the wrong choice. TRIGGER when: _analysis/design.md exists and is ready to promote, running `/synthesize-decisions` (Claude Code) or `@synthesize-decisions` (Cursor) after `/mine-design` (Claude Code) or `@mine-design` (Cursor), extracting design-decision practices from a codebase snapshot analysis.

Promote qualifying incident clusters from git history into reusable practice skills. TRIGGER when: _analysis/incidents.md exists and is ready to promote, running `/synthesize` (Claude Code) or `@synthesize` (Cursor) after `/mine-history` (Claude Code) or `@mine-history` (Cursor), extracting failure-mode practices from a mined repository.

Post-extraction quality pass over _analysis/extracted_skills_history/ and _analysis/extracted_skills_codebase/ — applies three tests to every skill and proposes keep, delete, merge, or rewrite. TRIGGER when: finishing `/extract-skills` (Claude Code) or `@extract-skills` (Cursor), `/synthesize` (Claude Code) or `@synthesize` (Cursor), or `/synthesize-decisions` (Claude Code) or `@synthesize-decisions` (Cursor), reviewing extracted skills before publishing, cleaning up duplicate or framework-specific skills, running `/curate-skills` (Claude Code) or `@curate-skills` (Cursor) after any extraction step.

Mine a repository's git history for institutional knowledge — failure modes, architectural decisions, and codebase directives. TRIGGER when: starting the history analysis pipeline, running `/mine-history` (Claude Code) or `@mine-history` (Cursor) before `/synthesize` (Claude Code) or `@synthesize` (Cursor), extracting practices from git history, looking for what has broken in this codebase and why.

Publish changed skills from _analysis/extracted_skills_codebase/ and _analysis/extracted_skills_history/ to the Prism registry via the Worker API with delta tracking. Only publishes skills whose content has changed since last publish unless --all is specified.

Mine this repository's git history for failure-mode practices and extract them as reusable skills, one step at a time, clearing context between steps (`/clear` in Claude Code, or clear context in Cursor) to prevent context blowup.