The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of script content found within a user's 'package.json' files to identify compatibility issues. This represents an attack surface where a malicious project file could attempt to influence the agent's behavior. However, the risk is mitigated by the skill's design, which requires user review and confirmation before any action is taken.
Ingestion points: Script values from 'package.json' files identified during the scan process in 'SKILL.md' (Step 1).
Boundary markers: The skill is explicitly instructed to present a summary of findings to the user (Step 3) and use 'AskUserQuestion' to confirm which categories to fix (Step 4) and which style to apply (Step 5) before making changes.
Capability inventory: The skill has the ability to modify local project files using the 'Edit' tool and execute shell commands via 'Bash' for dependency installation.
Sanitization: Content analysis is performed by a dedicated local script ('scripts/scan.mjs') using specific regular expression patterns.
[COMMAND_EXECUTION]: The skill executes a bundled Node.js script ('scan.mjs') to perform its analysis and uses the 'Bash' tool to run standard 'npm install' commands for adding necessary development dependencies to the user's project.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of well-known and trusted cross-platform utility packages ('cross-env', 'shx', and 'glob') from the official npm registry. It also provides instructions for installing Microsoft-provided system tools and the Python runtime via the 'winget' package manager for native dependency support.

windowsfy