reference-project-synthesizer
Pass
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface.
- Ingestion points: The skill reads 'paper cards' from the
reference/cards/directory (SKILL.md). These cards contain information extracted from external academic papers, which are untrusted third-party sources. - Boundary markers: The instructions do not define any specific delimiters or 'ignore' instructions to prevent the agent from treating data within a paper card as active instructions.
- Capability inventory: The skill has access to the
Bashtool and is instructed to write conclusions to project memory and code-related planning files likecode/.agent/benchmark-plan.md(SKILL.md, Step 5). - Sanitization: There is no mention of sanitizing or validating the content of the paper cards before they are used to influence the agent's logic or project updates.
Audit Metadata