The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The tool performs automated extraction of session cookies from the local browser. This involves reading sensitive browser database files to retrieve active session tokens, which is a credential harvesting pattern that bypasses secure authentication methods like OAuth.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of an external third-party package twitter-cli via the uv tool. This downloads and executes code from a remote source that is not verified in the context of this skill.
[COMMAND_EXECUTION]: The skill operates by executing shell commands (e.g., twitter search, twitter post) and requires the agent to modify the system's PATH environment variable.
[DATA_EXFILTRATION]: The combination of harvested credentials (cookies) and network-facing operations (posting to Twitter) creates a high risk of sensitive data exposure or session hijacking.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection. 1. Ingestion points: Untrusted data enters the agent context via twitter search and twitter feed in SKILL.md. 2. Boundary markers: The instructions lack delimiters or warnings to ignore instructions inside retrieved tweets. 3. Capability inventory: The skill allows subprocess calls (twitter post, twitter delete) and file writes (-o/--output). 4. Sanitization: There is no evidence of sanitization or validation of the external content before processing.

twitter-cli