interview
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input during the interview phase to generate a build sheet configuration.
- Ingestion points: User responses provided to the
AskUserQuestiontool inSKILL.mdare passed as CLI arguments (--job,--dod,--inputs, etc.) toscripts/interview_planner.py. - Boundary markers: Absent. User strings are interpolated directly into JSON fields for the agent's system prompt and rubric without delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill's scripts (
scripts/interview_planner.pyandscripts/build_sheet_builder.py) usepathlib.Path.write_text()to create and modify local JSON files (plan.jsonandbuild-sheet.json). No network or arbitrary code execution capabilities were found. - Sanitization: Minimal. The scripts perform basic string splitting and slugification, but do not escape or validate the content of the user-supplied job description or rubric items, allowing malicious instructions to be persisted in the resulting build sheet.
Audit Metadata