agents-md-improver
Pass
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it reads and incorporates content from potentially untrusted repo-local files (CLAUDE.md, GEMINI.md) into the agent's instructions. This risk is mitigated by the requirement for manual user approval of all proposed changes.
- Ingestion points: Reads instructions from AGENTS.md, CLAUDE.md, .claude/CLAUDE.md, and GEMINI.md within the repository scope (SKILL.md).
- Boundary markers: No explicit delimiters are defined to separate untrusted file content from instructions, but the skill implements a 'propose-first' workflow that serves as a manual validation step.
- Capability inventory: Possesses capabilities to read, write, and delete local markdown files, as well as create symbolic links.
- Sanitization: Does not explicitly sanitize or validate the content of the instruction files before processing.
- [COMMAND_EXECUTION]: The skill can trigger file system modifications, including the deletion of instruction files and the creation of symbolic links (e.g., linking CLAUDE.md to AGENTS.md). These operations are strictly constrained to the local repository and are protected by mandatory user confirmation prompts before any action is taken.
Audit Metadata