The Agent Skills Directory

[SAFE]: The skill functions as a local diagnostic utility to audit agent configurations, project structure, and maintainability metrics.
[DATA_EXPOSURE]: The skill accesses local configuration files (e.g., settings.local.json, CLAUDE.md) and conversation history (.jsonl logs) to perform its health check. This access is necessary for the audit and stays within the local environment.
[SAFE]: The data collection script check-agent-context.sh contains explicit redaction logic (using SENSITIVE_RE) to ensure that API keys, tokens, and secrets found in configurations are replaced with [REDACTED] before being processed by the agent.
[PROMPT_INJECTION]: The skill mitigates indirect prompt injection by providing explicit instructions to its sub-agents (inspector-context.md) to treat project data as untrusted input and to ignore any instructions embedded within the analyzed content.
[EXTERNAL_DOWNLOADS]: The skill includes instructions and logic to resolve its own installation path using the author's repository (tw93/Waza), which is consistent with the vendor's identity and does not pose a security risk.

health