Bluesky Search
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- Indirect Prompt Injection (LOW): The skill fetches untrusted content from the Bluesky network, which could contain malicious instructions designed to manipulate the AI agent.
- Ingestion points: External content is fetched in
scripts/search.pyfrom thepublic.api.bsky.appAPI. - Boundary markers: Absent; search results and post texts are displayed to the agent without explicit delimiters.
- Capability inventory: Low risk because the skill lacks write or execute capabilities (no
subprocess,eval, or file write operations). - Sanitization: None; the script relies on standard JSON parsing and simple string formatting.
- Data Exposure & Exfiltration (LOW): The script initiates network connections to a non-whitelisted domain.
- Evidence:
scripts/search.pymakesurllib.requestcalls tohttps://public.api.bsky.app, which is not on the predefined list of trusted domains.
Audit Metadata