The Agent Skills Directory

[SAFE]: The skill implements a structured workflow for documentation management that emphasizes user oversight and verification. It uses standard system tools like find and grep within appropriate scopes and enforces a strict 'report-before-edit' policy to ensure all changes are visible and confirmed by the user.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted external documentation files which could contain instructions intended to influence the agent. This is a characteristic of the tool's primary function and is mitigated by the requirement for manual user confirmation of all proposed changes.
Ingestion points: Discovery and reading of repository memory files such as CLAUDE.md, .claude.md, and .claude.local.md (SKILL.md, Phase 1 & 2).
Boundary markers: The instructions require the agent to generate a detailed quality report and present diffs for user approval before modifying any files (SKILL.md, Phase 3 & 4).
Capability inventory: The skill uses Read, Glob, Grep, Bash, and Edit tools to perform its auditing and documentation maintenance tasks.
Sanitization: The skill relies on manual user validation of proposed diff output rather than automated sanitization of the input documentation content.

memory-md-management