The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from the Bluesky social network, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Untrusted data enters the agent's context through the fetch_feed and search_feed functions in scripts/bluesky-reader.py, which retrieve posts from the public.api.bsky.app endpoint.
Boundary markers: The FeedResult.format_text method in scripts/bluesky-reader.py uses basic separators (---) between posts but does not include explicit instructions for the agent to ignore any embedded commands or prompts within the content.
Capability inventory: The skill uses the Bash tool to execute a Python script (SKILL.md) that possesses network access capabilities via urllib.request (scripts/bluesky-reader.py).
Sanitization: While the script parses structured JSON, it does not sanitize or filter the text content of the posts for prompt injection patterns before returning them to the agent.

bluesky-reader