The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a standalone shell script, find-polluter.sh, which performs directory traversal using find and automatically executes npm test on files matching a user-provided pattern.
[DATA_EXPOSURE]: The diagnostic instructions in SKILL.md provide examples for accessing and logging system state, including macOS keychain paths and security identities using the security command-line tool, as well as checking if specific environment variables like IDENTITY are set.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze potentially untrusted data such as error messages, stack traces, and logs. Evidence Chain: (1) Ingestion points: Error messages, multi-component logs, and git diffs. (2) Boundary markers: None defined to isolate data from instructions. (3) Capability inventory: Shell execution, system command execution, and file system modification. (4) Sanitization: No sanitization of external data is specified.
[DYNAMIC_EXECUTION]: The systematic debugging process encourages the agent to generate and inject custom diagnostic instrumentation and logging code into the application's source files to capture data at component boundaries.

systematic-debugging