hld-reviewer
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external design documents, such as HLDs and PRDs, as its primary input for review tasks. This creates an indirect prompt injection surface where instructions hidden within the analyzed documents could attempt to manipulate the agent's behavior. \n
- Ingestion points: The skill reads user-submitted HLD and PRD files as specified in
SKILL.mdandreferences/drift-detection-guide.md.\n - Boundary markers: The skill incorporates explicit boundaries defined in
references/review-boundaries.mdand requires the use of structured output templates (e.g.,references/report-templates.md) to maintain consistency.\n - Capability inventory: The agent can execute local Python scripts (
trace_lint.pyandtrace_build_rtm.py) to process the data.\n - Sanitization: The skill relies on automated linting scripts to validate the structure of the input data, which provides a layer of validation against malformed content.\n- [DYNAMIC_EXECUTION]: The skill invokes local utility scripts using
python3to automate the verification of requirement traceability. \n - Evidence:
SKILL.mdcontains instructions to runpython3 "$TESTANY_ENG_ROOT/scripts/trace_lint.py"andpython3 "$TESTANY_ENG_ROOT/scripts/trace_build_rtm.py".\n - Context: These scripts are part of the environment's internal toolchain (located under
$TESTANY_ENG_ROOT) and are used for their intended purpose of linting and building traceability matrices from the provided documents.
Audit Metadata