Audit prose artifacts for ambiguity and gaps across four subtypes.
npx skills add https://github.com/zhixuan312/multi-model-agent --skill mma-audit
Please help me install this Agent Skill. Skill: mma-audit Source: https://github.com/zhixuan312/multi-model-agent/tree/main/packages/server/src/skills/mma-audit Command: npx skills add https://github.com/zhixuan312/multi-model-agent --skill mma-audit
This Skill enables structured, cross-agent auditing of prose artifacts, code-execution plans, and SKILL.md to surface ambiguity, gaps, and mismatches before they propagate.
Post an audit request with subtype: default for a prose artifact to check coherence.
High-intent search queries and answers about installing and using this skill.
Auditing design docs involves performing a structured review to surface ambiguity and gaps in prose artifacts. It applies four subtypes: default, plan, spec, and skill, ensuring criteria-driven quality across documentation before mismatches propagate.
Yes, you can check code-execution plans against the codebase using the plan audit subtype. It performs plan-vs-codebase checks to surface mismatches and gaps between proposed execution steps and the actual codebase implementation.
The best way to validate requirement specs is using the spec audit subtype with context blocks. It associates upstream specs via SPEC COVERAGE context blocks to align requirements across rounds and validate requirement-prose accuracy.
To improve SKILL.md readability, apply the skill audit subtype. It performs a structured audit on SKILL.md files to surface readability gaps and ensure criteria-driven quality across cross-agent documentation and playbooks.
Yes, the audit process supports iterative improvements and delta tracking by integrating into MMA workflows via mma-context-blocks. This enables cross-agent rigor and tracks changes across multiple audit rounds.
Use the default audit subtype when checking general prose artifacts for coherence. It performs a structured audit to surface ambiguity and gaps in general documentation that does not fit the plan, spec, or skill subtypes.
wishx127Verify implementation completeness against change artifacts and source code evidence.
zhixuan312Register large documents once and reference them by ID in mma-* calls.
zhixuan312Dispatch ad-hoc tasks to workers and return structured per-task reports.
Nathanto-aiAnalyze systems to expose coherence gaps and risk across artifacts.
