portfolio-deal-linker-skill
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external communication and CRM platforms.
- Ingestion points: HubSpot deal activity/notes, Gmail thread content, and Apollo campaign data are ingested via MCP tools as described in
SKILL.md(Stages 1 and 2). - Boundary markers: There are no specific instructions to the agent to treat external data as untrusted or to ignore instructions embedded within deal notes or emails to prevent malicious content from influencing agent logic.
- Capability inventory: The skill possesses file-write capabilities (Stages 4 and 5) to update
portfolio.jsonl, weekly digests, and monthly reports based on the ingested data. - Sanitization: The instructions lack specifications for sanitizing or escaping external content before it is interpolated into portfolio records and metrics.
Audit Metadata