notion-knowledge-capture
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied conversations and notes to generate structured documentation in Notion. This creates a vulnerability where malicious instructions embedded in the source material could attempt to hijack the agent's behavior during the capture process.
- Ingestion points: Untrusted content enters the agent's context through conversation histories and supplied notes referenced in the main workflow (SKILL.md).
- Boundary markers: The prompt includes specific instructions to treat retrieved content as source material rather than instructions and explicitly forbids inventing owners, dates, or consensus.
- Capability inventory: The skill utilizes powerful tools including search, fetch, and the ability to create and update Notion pages and databases.
- Sanitization: The instructions rely on behavioral constraints to ensure factual fidelity, but there is no explicit mention of data escaping or sanitization of the interpolated content.
Audit Metadata