The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted data from external search results which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: search results (titles, snippets, and content) are fetched via httpx.get in scripts/searxng.py.
Boundary markers: Absent. The search results are output directly into the agent's context without delimiters or instructions to ignore embedded commands.
Capability inventory: The skill has network access and formatting capabilities via httpx and rich.
Sanitization: Absent. Result snippets are processed and displayed exactly as received from the SearXNG instance.
[Data Exposure & Exfiltration] (LOW): The script performs network requests to non-whitelisted, user-configurable domains and explicitly disables SSL security.
Evidence: scripts/searxng.py uses verify=False in the httpx.get request.
Risk: Disabling SSL verification (documented for local use) allows for Man-In-The-Middle (MITM) attacks if the SEARXNG_URL is set to a public instance over HTTPS.

searxng