IMPLEMENTAUDIT.md

What IMPLEMENTAUDIT.md does

The file skills/implementaudit/SKILL.md defines /implementaudit, a repo-generic method for turning audit findings, handoffs, checklists, reviews, goals, tasks, gaps and implementation plans into bounded, verified repository changes. Instead of trusting model self-assessment, it moves the definition of done into independently checkable evidence, using persistent state, executable checks, bounded decomposition and independent re-verification. A run counts as complete when it survives audit, and otherwise stops at an explicit audited handoff.

The method scales its ceremony to consequence: small reversible work stays compact, while deeper planning, durable run state or independent review activate when scope or risk calls for them. Its default authorization stance is no commit, no push, no tag, no release, no publication and no provenance, with each action needing separate explicit approval. The README says it is not an ungated autonomous build loop, release bot or package publisher, and that the .md in the repo name is branding, not a required root file.

Key features

• /implementaudit skill defined in skills/implementaudit/SKILL.md
• Completion judged by independently checkable evidence, not model self-assessment
• Persistent run state and bounded decomposition for long repository work
• Ceremony that scales with scope and risk
• Default stance of no commit, push, tag, release or publication without separate approval
• Greenfield and brownfield routing

When to use it

• Working through an AUDIT.md-style findings list with verified changes
• Planning and executing a repo task with independent re-verification
• Recovering proportionally when reality contradicts an implementation plan

Who it is for: Developers who direct coding agents on sustained repository work and want evidence-backed completion.

How it fits with Hermes Agent

The repository is tagged hermes-skill alongside claude-code-skill and codex-skill, so it targets several agents rather than Hermes alone.

FAQ