The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from various social media platforms (TikTok, Instagram, Reddit, etc.) through numerous proxy endpoints, creating a potential surface for indirect prompt injection.
Ingestion points: proxy-endpoints.md defines several endpoints for fetching videos, comments, posts, and profiles via gooseworks:data_call_provider.
Boundary markers: The skill does not instruct the agent to use delimiters or ignore instructions embedded within the retrieved data.
Capability inventory: The skill possesses file-write capabilities (logging to research/credits-log/<YYYY-MM>.md), network access via a proxy, and the ability to execute shell commands.
Sanitization: There are no instructions for sanitizing or escaping the content retrieved from external social media sources before it is processed or cited by the agent.
[COMMAND_EXECUTION]: The skill instructions include a suggestion to use shell commands for specific data retrieval tasks when other routes are unavailable.
Evidence: proxy-endpoints.md suggests using a shell curl command to fetch public data from itunes.apple.com if the standard proxy is not available. While the provided example is domain-specific, the use of shell commands represents an execution surface that should be monitored for broader exploitation.

spend-gate