claude-md-enhancer
Pass
Audited by Gen Agent Trust Hub on Feb 27, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes the content of local CLAUDE.md files which are considered untrusted data.
- Ingestion points: The skill ingests file content as strings in
analyzer.py(via theCLAUDEMDAnalyzerclass) andvalidator.py(via theBestPracticesValidatorclass). - Boundary markers: There are no explicit boundary markers or isolation mechanisms in the code to distinguish between project data and agent instructions.
- Capability inventory: The skill is granted permissions to
Read,Write,Edit, and execute specificBashcommands (ls,find,git). - Sanitization: No sanitization or filtering is performed on the ingested content to prevent embedded malicious instructions from influencing the agent's behavior.
- [COMMAND_EXECUTION]: The skill requests permissions to execute bash commands (
ls,find,git) for the purpose of exploring the repository structure during the initialization workflow. These commands are restricted using wildcards (e.g.,Bash(git:*)) to limit the scope of execution as a security precaution.
Audit Metadata