The Agent Skills Directory

[COMMAND_EXECUTION]: The find-polluter.sh script executes the npm test command on files matching a user-provided pattern. The SKILL.md and root-cause-tracing.md files also suggest using git, codesign, and security commands for diagnostic evidence gathering.
[PROMPT_INJECTION]: The skill uses directive instructions to prioritize systematic investigation over immediate fixes, such as "ALWAYS find root cause before attempting fixes." This influences the agent's task-solving strategy but does not target safety guardrails.
[PROMPT_INJECTION]: The skill facilitates the processing of untrusted data such as logs and error messages, creating a surface for indirect instructions.
Ingestion points: Error messages, stack traces, and log data are ingested in SKILL.md (Phase 1) and root-cause-tracing.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the documentation examples.
Capability inventory: The skill uses npm test in find-polluter.sh and suggests several CLI tools for system inspection.
Sanitization: There is no evidence of sanitization or filtering of the ingested log and error data before analysis.

systematic-debugging