The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to extract and utilize sensitive authentication credentials. It explicitly references using cookies via flags like --auth-token and --ct0, and encourages pointing the tool to local browser profile directories (e.g., --chrome-profile-dir) to extract session data.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external software from third-party sources not included in the trusted vendors list. This includes the Homebrew formula steipete/tap/bird and the NPM package @steipete/bird.
[COMMAND_EXECUTION]: The skill relies on executing the bird binary on the host system. This binary performs filesystem operations (reading browser cookie databases) and network requests based on user-provided arguments.
[PROMPT_INJECTION]: The skill acts as a vector for indirect prompt injection by ingesting untrusted data from X/Twitter.
Ingestion points: Content is retrieved from the platform via commands like bird read, bird search, bird home, and bird mentions.
Boundary markers: The skill lacks explicit instructions or markers to distinguish between retrieved social media content and system instructions.
Capability inventory: The skill possesses the ability to post content (bird tweet), modify account state (bird follow), and access local browser data.
Sanitization: There is no evidence of sanitization or filtering of the content retrieved from external social media feeds before it is processed by the agent.

bird