The Agent Skills Directory

[COMMAND_EXECUTION]: Executes an inline Python script using a heredoc to create and update the local configuration file ~/.ouroboros/prefs.json. While the logic is limited to JSON manipulation, it involves direct script execution on the host system.
[COMMAND_EXECUTION]: Invokes the GitHub CLI (gh api) to perform authenticated requests. Specifically, it executes gh api -X PUT /user/starred/Q00/ouroboros to star a repository on the user's behalf.
[DATA_EXFILTRATION]: Performs an external write operation by interacting with the GitHub API to modify user account data (starring a repository). This uses the user's existing CLI authentication to interact with an external service controlled by the vendor.
[PROMPT_INJECTION]: Subject to indirect prompt injection (Category 8) because it extracts requirements and data from conversation history to generate structured YAML specifications.
Ingestion points: Reads interview session data and Q&A history from the conversation context.
Boundary markers: None identified; instructions do not specify delimiters or warnings to ignore embedded instructions in the interview data.
Capability inventory: File system writes (prefs.json), shell command execution (Python), and external API interaction (gh CLI).
Sanitization: No evidence of validation or sanitization of the interview results before they are processed into the Seed YAML output.

seed