trace-requirements-impact
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied 'Change Requests' which are considered untrusted external input and could contain malicious instructions designed to bypass project constraints.
- Ingestion points: Stakeholder change requests are ingested at the start of the analysis process as described in Step 1 of the 'Suggested Process' in
SKILL.md. - Boundary markers: The skill includes a specific security guardrail (ASI01 Goal Hijack) that instructs the agent to treat change request content as untrusted input and ignore any instructions to skip analysis or safety filters.
- Capability inventory: The skill utilizes several high-capability tools including
execute_command,read_file,write_file,edit_file,search_code,run_tests,run_linter, andrun_buildto perform impact analysis. - Sanitization: The skill relies on enforced procedural rules (e.g., BFS traversal up to 3 hops) to ensure a complete and consistent analysis is performed regardless of the wording inside the change request.
Audit Metadata