systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides utility scripts and command-line examples meant for local diagnostics and automated testing during debugging sessions.
- Evidence: The 'find-polluter.sh' script executes 'npm test' on files matching a specified pattern to identify state-polluting tests.
- Evidence: 'SKILL.md' suggests executing diagnostic commands such as 'codesign', 'security list-keychains', and environment variable checks ('env | grep') to troubleshoot multi-component build systems.
- [INDIRECT_PROMPT_INJECTION]: The skill creates a potential vulnerability surface by instructing the agent to ingest and analyze external data such as error messages and logs.
- Ingestion points: Error messages, stack traces, logs from multi-component systems, and git diffs (SKILL.md, root-cause-tracing.md).
- Boundary markers: The instructions do not specify the use of delimiters or 'ignore' directives when processing external log data.
- Capability inventory: The skill assumes the ability to execute local testing frameworks (npm), code signing tools, and system diagnostic utilities (SKILL.md, find-polluter.sh).
- Sanitization: There is no mention of sanitizing or escaping the content of processed logs before they are analyzed by the agent.
Audit Metadata