config-health
Warn
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: MEDIUMPERSISTENCEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PERSISTENCE]: The skill documentation guides the user to modify the agent's configuration file (settings.json) to register hooks for SessionStart and PreToolUse. These hooks ensure that the config-health.py script is executed automatically every time a session begins or when specific tools (Edit/Write) are matched, creating a persistent execution mechanism.
- [EXTERNAL_DOWNLOADS]: The primary installation method uses
npx ecc-universal install, which fetches and executes code from the npm registry at runtime. The package 'ecc-universal' is not a well-known or trusted service, introducing a supply chain risk if the package is compromised or malicious. - [COMMAND_EXECUTION]: The skill operates by triggering shell commands (e.g.,
python3 ~/.claude/scripts/config-health.py) through agent hooks. This automated execution of local scripts depends on the integrity of the installation process and the security of the script files provided. - [INDIRECT_PROMPT_INJECTION]: The skill processes project-level data that could be modified by external actors to influence the agent's startup audit.
- Ingestion points: The script reads and parses settings.json, .claude/settings.local.json, CLAUDE.md, and rule-health.jsonl.
- Boundary markers: No specific boundary markers or safety instructions are mentioned to prevent the agent from being misled by content within these files.
- Capability inventory: The skill possesses the ability to execute shell commands and audit filesystem paths across multiple directory roots.
- Sanitization: While the documentation outlines logic for handling shell-style quoting and path resolution, it lacks explicit sanitization to prevent malicious data in the configuration files from affecting the agent's behavior during the audit.
Audit Metadata