product-skills
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-generated data in the form of discovery logs and Opportunity Solution Trees to calculate health metrics and structural compliance, which constitutes a potential indirect prompt injection surface.\n
- Ingestion points: Local JSON files such as
discovery_log.jsonandost.jsonare read by analysis scripts (scripts/discovery_cadence_tracker.py,scripts/ost_linter.py).\n - Boundary markers: The data is structured as JSON, but there are no explicit delimiters to distinguish data from instructions within the string fields.\n
- Capability inventory: The skill is capable of executing local Python scripts and generating plans through a referenced repository-wide agent harness.\n
- Sanitization: Input data is parsed using standard JSON libraries and validated against specific structural rules defined in the linter and tracker scripts.\n- [COMMAND_EXECUTION]: The skill orchestrates behavior by executing local Python scripts using user-supplied goal descriptions as command-line arguments.\n
- Evidence:
SKILL.mdcontains instructions for the agent to run commands such aspython3 scripts/product_goal_router.py --text "<the goal>"to classify tasks.
Audit Metadata