The Agent Skills Directory

[SAFE]: The skill is entirely instructional, providing a reference for the pnpm package manager. It describes standard CLI usage, configuration via pnpm-workspace.yaml, and store management without any suspicious code or directives.
[REMOTE_CODE_EXECUTION]: The documentation mentions standard pnpm features for executing remote or local packages, such as pnpm dlx, pnx, and pnpm exec. These are legitimate tool capabilities and no malicious execution patterns (like piping external shell scripts to bash) are present in the instructions.
[SUPPLY_CHAIN_SECURITY]: The skill emphasizes pnpm's supply-chain security features, specifically documenting the allowBuilds configuration (which blocks unreviewed lifecycle scripts), minimumReleaseAge (to avoid day-zero malicious releases), and trustPolicy settings. These instructions guide the user/agent toward secure package management practices.
[INDIRECT_PROMPT_INJECTION]: The skill defines how the agent should ingest and interpret project files such as package.json, pnpm-workspace.yaml, and .pnpmfile.mjs. While these are external inputs, the skill provides standard interpretation logic for dependency management and does not include patterns that would lead to instruction override or exploitation.
[DYNAMIC_EXECUTION]: The skill mentions the use of .pnpmfile.mjs for hooks. This is a standard extensibility mechanism in pnpm for programmatically modifying manifests or configurations. The provided examples are benign (e.g., adding peer dependencies or pinning versions).

pnpm