reference-project-synthesizer

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest "source cards" (summaries of external documents like papers, collaborator docs, and Markdown notes) and use them to update project memory.
  • Ingestion points: The skill reads external data from paths such as reference/cards/<source-id>.md and various project memory files in memory/ or code/.agent/.
  • Boundary markers: There are no explicit instructions or delimiters in the analyzed files to separate the agent's internal logic from potentially untrusted content within the source cards.
  • Capability inventory: The skill is configured with powerful capabilities including Write, Edit, and Bash tools.
  • Sanitization: The provided instructions do not specify any validation or sanitization of the source card content before it is processed and used to modify the project's durable memory (e.g., evidence-board.md, action-board.md).
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 06:18 PM
Security Audit — agent-trust-hub — reference-project-synthesizer