The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a utility script find-polluter.sh for automating the search for tests that leave behind unwanted side effects. It also provides examples of using diagnostic commands like codesign and security for troubleshooting build and signing issues.
[INDIRECT_PROMPT_INJECTION]: The skill instructs agents on how to process external inputs such as bug reports and error logs. The framework's rigid 4-phase structure guides the agent to rely on evidence and systematic verification. Ingestion points: Bug reports and test failure output (SKILL.md). Boundary markers: None explicitly mentioned. Capability inventory: Shell command execution via find-polluter.sh and agent tools. Sanitization: The find-polluter.sh script uses quoted variables to prevent simple shell injection when handling file paths.
[DATA_EXPOSURE]: Documentation within the skill demonstrates safe practices for verifying the presence of secrets in an environment (e.g., using ${VARIABLE:+SET} in shell scripts) without revealing their actual values.

systematic-debugging