The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/bin/npm-upgrade.mjs uses spawnSync to execute shell commands including git, npm, pnpm, yarn, and gh for repository management, package installation, and build verification.- [EXTERNAL_DOWNLOADS]: The skill invokes npx, pnpm dlx, or yarn dlx to download and execute the npm-check-updates utility from the public npm registry. It also performs standard package installations which download dependencies from the registry.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from the npm registry which could contain malicious instructions designed to influence the agent's PR generation.
Ingestion points: The fetchPackageMetadata function in scripts/bin/npm-upgrade.mjs retrieves package descriptions and URLs from https://registry.npmjs.org/ and stores them in .npm_upgrade/changelog_diffs.json.
Boundary markers: The instructions in SKILL.md lack explicit delimiters or instructions to ignore embedded commands within the fetched metadata when generating summaries.
Capability inventory: The agent has the capability to execute shell commands (via the helper script) and create GitHub Pull Requests using the gh CLI.
Sanitization: There is no evidence of sanitization for the metadata fetched from the npm registry before it is presented to the AI for summarization.

npm-upgrade