The Agent Skills Directory

[CREDENTIALS_UNSAFE]: A hardcoded Google API key was found in the application source code.
Evidence: index.js contains this.apiKey = "AIzaSyCKWmPmAkZWvI2KiblawWPUESyCp9dEjk0";.
[COMMAND_EXECUTION]: The skill's instructions recommend bypassing security boundaries during command execution.
Evidence: SKILL.md promotes the use of the --yolo flag for the Codex CLI, which it describes as having "NO sandbox, NO approvals".
[EXTERNAL_DOWNLOADS]: The documentation directs users to install external packages from third-party repositories.
Evidence: SKILL.md provides instructions to run npm install -g @mariozechner/pi-coding-agent.
[PROMPT_INJECTION]: The skill processes untrusted external data, such as Pull Request content, which creates an attack surface for indirect prompt injection.
Ingestion points: Pull Request descriptions and git diffs are ingested for analysis in SKILL.md (e.g., codex review).
Boundary markers: No delimiters or instructions to ignore embedded commands are provided to the agent.
Capability inventory: The skill has extensive shell execution capabilities via the bash tool, including file system access and the ability to run external AI agents.
Sanitization: No sanitization or validation of the external content is performed before it is processed by the agents.

coding-agent