devrel-radar
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from both external Markdown files and user interview responses to generate a personalized radar report.
- Ingestion points: The skill reads source metadata and practitioner details from
references/sources.mdandreferences/people-to-follow.md, and integrates direct user input from the interview workflow. - Boundary markers: The instructions lack specific delimiters or escaping mechanisms to isolate external content from the agent's output instructions.
- Capability inventory: The skill performs file read operations and network status checks (probes).
- Sanitization: No explicit validation or sanitization is performed on the descriptions or names retrieved from external source files before they are presented to the user.
- [DATA_EXFILTRATION]: The skill performs network operations to verify the status of external resources.
- Evidence: The 'Freshness and ownership check' logic requires the agent to fetch canonical URLs to identify dated items (episodes, posts, events) and classify links as active, stale, or dead. These network probes are restricted to status verification and do not involve the transmission of sensitive local data.
Audit Metadata