The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The references/tool-installer.md script downloads the nvm installation script from a public GitHub repository (https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh) and executes it directly through bash. This is the standard installation method for nvm, a widely recognized Node.js version manager, but constitutes remote code execution.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads and analyzes numerous files from potentially untrusted repositories, including CLAUDE.md, .claude/rules/*.md, and .claude/settings.json. A malicious repository could embed instructions in these files to manipulate the auditor agent's behavior. * Ingestion points: Reads project metadata, source code, CI/CD workflows, and AI harness configuration files. * Boundary markers: There are no explicit boundary markers or specific warnings for the agent to ignore instructions found within the data files. * Capability inventory: The skill is permitted to use Bash, Edit, Write, Agent, and Task tools, allowing significant actions if the agent is compromised. * Sanitization: The skill does not perform sanitization or validation of the content ingested from the project files before processing.
[COMMAND_EXECUTION]: The skill performs extensive command execution as part of its auditing process, including managing Node.js versions with nvm, installing dependencies via npm, yarn, or pnpm, and running analysis commands like find, grep, wc, tsc, and ng. These actions are necessary for the skill's intended purpose of auditing a frontend application.
[EXTERNAL_DOWNLOADS]: The skill triggers the download of the nvm installer and project dependencies from public registries (npm, yarn, pnpm). These downloads target well-known development infrastructure and are required for environment setup and analysis.

angular-health-audit