The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it is designed to ingest and process untrusted data from various social media platforms.
Ingestion points: The references/listening.md file describes workflows for fetching real-time content from Reddit, Hacker News, and Bluesky using API endpoints.
Boundary markers: While the skill advises the user to review drafts before posting, it lacks explicit prompt-level delimiters or instructions to ignore potential commands embedded in the fetched data.
Capability inventory: The skill utilizes shell commands (curl) and browser tools (dev-browser) to interact with external network resources.
Sanitization: No procedures are defined for sanitizing or escaping content retrieved from external sources before it is interpreted by the agent.
[COMMAND_EXECUTION]: The skill provides several shell command 'recipes' in references/listening.md that include variable placeholders for search keywords and channel IDs.
Evidence: The recipes for searching Reddit, Hacker News, and Bluesky (e.g., curl -s ... "query=KEYWORD" | jq) interpolate variables directly into shell commands. If these variables are populated with unsanitized user input or data from untrusted sources, it could lead to command injection if the agent executes them in a standard shell environment.
[EXTERNAL_DOWNLOADS]: The skill fetches data from several well-known external services to support its social listening features.
Evidence: Instructions in references/listening.md detail how to use curl to retrieve JSON and XML data from Reddit (reddit.com), Algolia/Hacker News (hn.algolia.com), Bluesky (public.api.bsky.app), and YouTube (youtube.com). These are categorized as safe, well-known services used for their intended purpose.

social