The Agent Skills Directory

[Command Execution] (LOW): The find-polluter.sh script executes arbitrary local test files using npm test. While this is standard for a debugging utility, it allows for the execution of code within the environment.
[Indirect Prompt Injection] (LOW): The skill instructs the agent to ingest and analyze external data like error messages, stack traces, and logs. This creates a surface for indirect prompt injection if the ingested data contains instructions designed to manipulate the agent's logic.
Ingestion points: SKILL.md Phase 1 (Error messages, stack traces), Phase 4 (Test results).
Boundary markers: Absent. The skill does not define specific delimiters for separating user data from system instructions.
Capability inventory: Local command execution (npm test, security, codesign) as described in find-polluter.sh and SKILL.md examples.
Sanitization: Absent. The skill does not suggest sanitizing or escaping content from logs or error messages before processing.
[Data Exposure] (SAFE): The provided code examples in SKILL.md demonstrate best practices for checking the existence of secrets (e.g., using ${IDENTITY:+SET}) without revealing the actual secret values.

systematic-debugging