The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a bash script find-polluter.sh that automates the execution of local tests using npm test to isolate code responsible for state pollution.
[COMMAND_EXECUTION]: The debugging methodology encourages the use of system commands such as env, security list-keychains, and codesign to inspect environment variables and security configurations during failure investigation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection vulnerabilities. 1. Ingestion points: The agent is instructed to ingest and analyze external error messages, stack traces, and component logs in Phase 1. 2. Boundary markers: No delimiters or explicit warnings to ignore instructions within the logs are present. 3. Capability inventory: The agent has the capability to execute shell commands and run local test suites. 4. Sanitization: The process does not specify sanitization or validation steps for the diagnostic data before it is processed by the agent.

systematic-debugging