By name: PromptAnalysis version: 0.1.0 description: "Validate and minimize prompts — provenance, targeting, staleness, redundancy, ablation testing, perplexity scoring. USE WHEN audit rules, check provenance, minimize prompts, prompt cleanup, validate targeting, find redundant rules, prompt audit, prompt analysis, stale rules."
PromptAnalysis
Analyze rules, skills, and agents for minimum viable prompt compliance (MVPR-0001). Each concern is a subskill with its own detection patterns and output format.
Subskill Routing
| Keyword | Subskill | What it does |
|---|---|---|
| "scan", "audit", "check", "all" | Scan | Run all subskills on a module |
| "provenance", "sources", "refs" | @Provenance.md | Check external source citations |
| "target", "targeting", "provider" | @Targeting.md | Check model/provider targeting correctness |
| "stale", "staleness", "deprecated" | @Staleness.md | Check for outdated references |
| "cleanup", "minimize", "fix" | Cleanup | Scan + auto-fix with confirmation |
| single file path | Single | Full analysis of one file with all subskills |
External Tool Subskills
| Keyword | Subskill | Requires |
|---|---|---|
| "deep", "test", "ablation" | @PromptFoo.md | npm install -g promptfoo + API keys |
| "compress", "perplexity" | @LLMLingua.md | pip install llmlingua |
| "benchmark", "capability" | @ArtificialAnalysis.md | API key from artificialanalysis.ai |
Scan Mode
Run all static subskills (Provenance + Targeting + Staleness) on every rule in a module:
/PromptAnalysis scan Modules/forge-core/rules
Scan collects results from each subskill and presents a unified report grouped by status. No API calls, no model queries — purely static.
Single File Mode
All subskills on one file:
/PromptAnalysis Modules/forge-core/rules/NoHeredoc.md
Shows full content with inline annotations, current targeting, qualifier variants, and a recommendation.
Cleanup Mode
Scan + apply safe structural fixes with confirmation:
/PromptAnalysis cleanup Modules/forge-core/rules
For each fixable issue, present via AskUserQuestion. Only structural fixes (add targets, move to qualifier dir). Never delete rules or change content.
Output Format
Follow @Output.md for the ASCII dashboard layout, card types, optional gauges, and batch decision flow.