The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for executing shell commands using the bash tool, specifically promoting the use of the --yolo flag for the Codex CLI. This flag is explicitly described as disabling sandboxing and approvals, which facilitates autonomous execution of code changes without human-in-the-loop verification. Additionally, the bash tool documentation within the skill mentions an elevated parameter for running commands on the host system instead of a sandbox.
[PROMPT_INJECTION]: The skill describes workflows for reviewing pull requests and fixing issues by cloning external GitHub repositories. This creates an indirect prompt injection surface where malicious instructions or adversarial patterns embedded in the external code or PR metadata could influence the behavior of the coding agent or the host agent.
Ingestion points: External data enters the agent context via git clone, gh pr checkout, and git fetch operations described in the SKILL.md files.
Boundary markers: The instructions do not provide explicit boundary markers or warnings to the coding agents to treat the external repository content as untrusted data.
Capability inventory: The skill has access to the bash tool (with potential host access via the elevated parameter) and the process tool for session manipulation (log, write, submit).
Sanitization: There is no evidence of sanitization, validation, or filtering of the external content before it is processed by the coding agents.
[EXTERNAL_DOWNLOADS]: The skill instructions include commands to download and install software from external sources, such as npm install -g @mariozechner/pi-coding-agent and cloning repositories from GitHub. While these target well-known services, they involve the download and execution of remote content.

coding-agent