The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands based on file paths and configuration values retrieved from the local project environment. This introduces a risk of command injection if filenames or project settings are maliciously crafted.
[REMOTE_CODE_EXECUTION]: The skill automatically identifies and runs scripts defined within package.json (such as lint, typecheck, and format scripts). This enables arbitrary code execution if an attacker provides a project containing malicious script definitions, posing a significant supply chain risk.
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection through untrusted data within the codebase. Ingestion points: Reads tool definitions and lifecycle scripts from package.json, pyproject.toml, and other local configuration files. Boundary markers: None; the skill directly interpolates external configuration into shell command strings. Capability inventory: Broad shell execution capabilities across multiple environments (Node.js, Python, Go, Rust). Sanitization: No validation or escaping of the script content found in configuration files is performed before execution.

preflight-checks