The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to access sensitive credentials during diagnostic phases. In SKILL.md, it provides examples of commands to inspect the macOS keychain (security list-keychains) and environment variables containing identity tokens (env | grep IDENTITY). \n- [DATA_EXFILTRATION]: The instructions encourage logging sensitive data at component boundaries. This practice can result in secrets or cryptographic keys being exposed in agent output or logs during the "gathering evidence" phase. \n- [COMMAND_EXECUTION]: The find-polluter.sh utility script executes npm test on files discovered through a user-specified search pattern. This enables the execution of arbitrary code if a malicious file matches the provided pattern. \n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted inputs (error messages, stack traces, and logs) without using delimiters or instructions to ignore embedded commands. \n
Ingestion points: Error messages and system logs analyzed in Phase 1 (SKILL.md). \n
Boundary markers: Absent. \n
Capability inventory: Shell command execution via various debugging and build tools. \n
Sanitization: Absent.

systematic-debugging