systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
find-polluter.shscript executes a user-specified command (e.g., a test runner) across multiple files to identify the source of test pollution. It includes security checks to ensure test files reside within the current workspace and are not symlinks resolving to external locations. - [DYNAMIC_EXECUTION]: The
waitForutility inscripts/condition-based-waiting-example.tsexecutes a provided probe function repeatedly at runtime until a condition is met or a timeout occurs. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data, such as logs and error outputs, during the debugging process. It includes explicit instructions to redact secrets and personal data, and to sanitize diagnostics to prevent data exposure or manipulation.
- Ingestion points: Error logs, terminal output, and source code files as described in
SKILL.mdandreferences/root-cause-tracing.md. - Boundary markers: The skill relies on instructional constraints for the agent to maintain safety boundaries.
- Capability inventory: Execution of shell commands via
find-polluter.shand function execution viawaitFor. - Sanitization: Instructions recommend redacting credentials and environment variables from logs.
Audit Metadata