The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a template for ralph.sh which executes the claude CLI with the --dangerously-skip-permissions flag. This flag bypasses built-in safety prompts, enabling the AI to perform autonomous shell command execution. The setup instructions also include making the script executable via chmod +x.
[REMOTE_CODE_EXECUTION]: The autonomous runner implements and executes code (e.g., yarn typecheck) based on instructions found in external PRD files, facilitating the execution of AI-generated logic in a continuous loop.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted PRD files.
Ingestion points: Data from markdown PRD files is converted into prd.json and then passed as context to the Claude CLI using the @ file reference syntax.
Boundary markers: No delimiters or instructions are used to distinguish between system instructions and data ingested from external files.
Capability inventory: The AI is granted the ability to execute shell commands (with permissions skipped), write to the filesystem, and perform git operations.
Sanitization: No sanitization or validation is applied to the content of the PRD files before it is processed by the autonomous runner.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the jq utility via the well-known Homebrew package manager if it is not already available on the system.

ralph