compose-outreach
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from Common Room (community posts, activity) and web search results (articles, news) to personalize outreach drafts. This ingestion of untrusted data creates a potential surface for indirect prompt injection.
- Ingestion points:
SKILL.mdStep 1 and Step 2 describe fetching target data and external hooks. - Boundary markers: The instructions lack specific delimiters or instructions to ignore commands hidden within retrieved data.
- Capability inventory: The skill uses MCP tools and web search but does not perform sensitive operations like file modification or exfiltration.
- Sanitization: No sanitization is specified for the ingested content.
Audit Metadata