The Agent Skills Directory

[COMMAND_EXECUTION]: The skill incorporates a utility script (find-polluter.sh) and various shell command examples in SKILL.md to help identify the source of test failures and verify environment configurations. These tools use standard commands such as npm test, find, env, and macOS security utilities for diagnostic purposes.
[INDIRECT_PROMPT_INJECTION]: The debugging workflow requires the agent to ingest and analyze external data like error logs, stack traces, and test outputs, which could serve as a vector for indirect prompt injection.
Ingestion points: Phase 1 of the debugging process in SKILL.md and the tracing instructions in root-cause-tracing.md direct the agent to examine error output and logs.
Boundary markers: The skill lacks explicit instructions to treat error messages as untrusted data or to use specific delimiters when processing them.
Capability inventory: The skill allows for command execution and file system access to facilitate the debugging process.
Sanitization: The instructions do not specify a requirement for sanitizing or escaping the technical data processed during investigation.

systematic-debugging