The Agent Skills Directory

COMMAND_EXECUTION (SAFE): The skill contains a utility script find-polluter.sh that automates test execution using npm test to identify state pollution. While this executes local code, it is a standard developer tool operation and does not involve remote scripts or untrusted data execution.
DATA_EXPOSURE (SAFE): The documentation includes examples of using macOS security commands (security list-keychains, security find-identity) for debugging code signing workflows. These are used as diagnostic examples and do not attempt to exfiltrate or store sensitive credentials.
INDIRECT PROMPT INJECTION (LOW): The skill's primary function is to ingest and analyze untrusted data such as error messages, stack traces, and test logs.
Ingestion points: Error messages, stack traces, and component logs processed during Phase 1 of the debugging framework.
Boundary markers: The skill encourages the use of explicit logging prefixes (e.g., DEBUG git init:) to distinguish diagnostic output.
Capability inventory: The skill utilizes local shell execution (via npm test and codesign) to verify fixes and gather environment data.
Sanitization: No explicit sanitization of error logs is mentioned, but the risk is mitigated as the analysis is intended for developer-controlled local environments.
BEST PRACTICE (SAFE): The script find-polluter.sh uses a for loop over the results of find, which may fail on filenames containing spaces. This is a minor reliability issue rather than a security vulnerability.

debugging