The Agent Skills Directory

[DATA_EXFILTRATION] (HIGH): The skill is designed to 'auto-detect' and extract sensitive authentication cookies (auth_token and ct0) from Chrome, Firefox, Edge, and Safari browsers on the host machine to facilitate 'zero-config' Twitter access.
[CREDENTIALS_UNSAFE] (HIGH): In the REST and ACP protocols, these extracted credentials are sent to a remote third-party server (https://xbirdapi.up.railway.app). While the documentation claims the server is 'stateless' and uses 'E2E encryption', the credentials leave the local security boundary to an unverified endpoint.
[COMMAND_EXECUTION] (HIGH): Setup instructions require users to run npx @checkra1n/xbird or bunx @checkra1n/xbird. This executes code from an npm package maintained by an untrusted author, allowing for arbitrary local command execution.
[EXTERNAL_DOWNLOADS] (MEDIUM): The installation process fetches and runs multiple packages from the npm registry (@checkra1n/xbird, @x402/fetch) which are not part of the trusted developer list.
[PROMPT_INJECTION] (LOW): The skill ingests untrusted data from the Twitter API (e.g., via search_tweets and get_mentions) and presents it to the agent, creating a surface for indirect prompt injection attacks.
Ingestion points: tools.md (search_tweets, get_mentions, get_home_timeline)
Boundary markers: None identified in the provided tool documentation.
Capability inventory: tools.md (post_tweet, update_profile, upload_media)
Sanitization: No evidence of sanitization for incoming tweet content.

xbird