The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill executes npx skills@latest, which downloads the latest version of the skills package from the npm registry. This is the intended behavior for the tool's functionality.
[COMMAND_EXECUTION]: The skill executes shell commands to find files (find), run the CLI tool (npx), and commit changes to the repository (git commit). These operations are limited to the project directories specified or found.
[INDIRECT_PROMPT_INJECTION]: The skill reads keys from skills-lock.json files and uses them as arguments in a shell command, creating a potential surface for command injection if the file content is untrusted.
Ingestion points: skills-lock.json file found on the local file system.
Boundary markers: Absent; there are no instructions to sanitize or escape the keys extracted from the JSON file before passing them to the shell.
Capability inventory: Includes shell command execution via npx, find, and git across the file system.
Sanitization: Absent; the agent is instructed to use the keys directly as skill names.

update-project-skills