The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a bash script (find-polluter.sh) that automates running npm test against files found via a user-defined search pattern. This facilitates the execution of local code during debugging.
[DATA_EXFILTRATION]: Example diagnostic commands in SKILL.md demonstrate accessing potentially sensitive system information, including macOS keychain configurations (security list-keychains), cryptographic identities (security find-identity), and environment variables (env).
[INDIRECT_PROMPT_INJECTION]: The debugging process relies on the ingestion of untrusted external data, such as stack traces, error messages, and git diffs.
Ingestion points: Phase 1 of SKILL.md mandates reading error messages, logs, and git diffs.
Boundary markers: No specific delimiters or safety instructions are provided to distinguish between system output and potentially malicious content within logs/diffs.
Capability inventory: The agent has capabilities to execute shell commands and run tests (e.g., find-polluter.sh, codesign examples).
Sanitization: No sanitization of the error messages or logs is performed before analysis.

systematic-debugging