The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses python3 -c in its setup instructions to extract configuration values and credentials from a local JSON file. This is used to set environment variables required for its operation.
[DATA_EXPOSURE]: The skill instructions read from ~/.gooseworks/credentials.json. This is a vendor-specific credential file used by the skill author (gooseworks-ai) to manage authentication for their own API services. The extracted information is used locally for authorization headers in API calls.
[DATA_INJECTION]: By fetching and searching web content, the skill introduces untrusted external data into the agent's context. This is an inherent surface for indirect prompt injection, typical for web-browsing capabilities. Documentation of this surface is as follows:
Ingestion points: The /search and /fetch endpoints in SKILL.md retrieve external web content.
Boundary markers: None present in the current instructions to delimit fetched content.
Capability inventory: The skill can execute shell commands (python3) and network requests (curl).
Sanitization: No explicit sanitization of fetched web content is described.

web-search-linkup