The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the retrieval of untrusted data from external social media sources (X/Twitter). This content, which includes posts and profile information, could contain hidden instructions or malicious payloads intended to influence the agent's behavior.
Ingestion points: Data enters the system context through various search and timeline endpoints, such as /api/search and /api/user/{id}/tweets, as described in SKILL.md.
Boundary markers: The instructions do not provide explicit delimiters or prompts to the agent to treat the fetched content as data rather than instructions.
Capability inventory: The skill provides access to wide-ranging social media data, which is then processed by the agent to fulfill user requests.
Sanitization: No filtering or sanitization of the retrieved social media content is specified in the documentation.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill involves network communication with an external host to retrieve data.
Evidence: The base URL https://twitter.fetcher.sh is used across all endpoints for API calls, as seen in SKILL.md and references/endpoints.md.
Context: The target domain is a vendor-owned resource associated with fetcher-sh and is necessary for the skill's intended functionality.

x-api