The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content fetched from the internet via search results, which creates a potential surface for indirect prompt injection attacks.
Ingestion points: The https://google.fetcher.sh/api/search endpoint retrieves external web content.
Boundary markers: The skill instructions do not specify any delimiters or safety prompts to isolate search result content.
Capability inventory: Depending on the agent's other skills, it may have capabilities for file manipulation or command execution that could be targeted by instructions hidden in search results.
Sanitization: No explicit sanitization or filtering of search result content is described.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill performs network operations to google.fetcher.sh to retrieve search data.
This communication is the primary purpose of the skill and is conducted over secure HTTPS.
The skill correctly instructs users to manage credentials via environment variables (FETCHER_API_KEY) and provides placeholders for configuration examples, avoiding the exposure of actual secrets.

google-search