audit-prototype
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from product requirement documents (PRD) and implementation logs in the project workspace. This creates a surface for indirect prompt injection if those files contain instructions designed to manipulate the agent's reasoning or reporting outputs.
- Ingestion points: Processes files such as .agents/flow/it_{iteration}PRD.json and .agents/flow/it{iteration}_progress.json.
- Boundary markers: The instructions do not include boundary markers or explicit warnings for the agent to ignore instructions embedded within the processed data files.
- Capability inventory: The skill has the ability to write files to the project directory and execute shell commands via the nvst utility.
- Sanitization: There is no evidence of sanitization, validation, or escaping of the data extracted from project files before it is interpolated into prompts or command arguments.
- [COMMAND_EXECUTION]: The skill invokes the nvst command-line tool to update technical debt records and write structured audit reports. This execution is performed using data generated from the analysis of external, potentially untrusted documents.
Audit Metadata