Аудит AI-экосистемы репозитория

★ 6.7 · vibe-coding

ai-ecosystem-audit is a Claude Code skill that runs a structured, read-only audit of repository-controlled AI assets when invoked via `/ecosystem-audit` or the `ecosystem-auditor` specialist. It inspects skills, agents, commands, and configs under `.claude/` and `.opencode/`, checking mirror parity by logical name and contract, reference integrity, mutation boundaries, and platform fit across `gc-platform` and `portable` relevance modes. The `--scope` flag narrows analysis to parity, skills, agents, commands, or config; `--strict` promotes unverified references and incomplete mirrors from risks to confirmed findings without converting inferences into facts. Every inventory assertion and finding is cited as `path:line` with a confidence level — confirmed, likely, or unknown. Output is a single Markdown document containing a parity inventory, findings table, purpose map, deliberate-differences register, risks, and a prioritized backlog — useful for teams auditing harness consistency across two AI ecosystems.