The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates the full deployment pipeline by executing shell commands for version control (git), project building (npm/pnpm/yarn), and deployment (wrangler, vercel). It also utilizes dynamic execution methods such as the source command to load environment variables and node -e to parse local configuration files. Additionally, it instructs the user to modify shell profile configurations like ~/.zshrc to persist deployment credentials.
[DATA_EXFILTRATION]: Accesses and reads sensitive environment variable files, including .env, .env.production, and .env.prd, to retrieve authentication tokens (e.g., CLOUDFLARE_API_TOKEN) required for deployment. This access is limited to standard local paths for legitimate deployment purposes.
[PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted data from the local repository, specifically through git diff output and the contents of package.json and wrangler.jsonc files to generate commit messages and determine deployment logic.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when interpolating local file data into the agent's context.
Capability inventory: The skill possesses significant capabilities, including the ability to run arbitrary shell commands via npm scripts, perform network operations via git push, and deploy code to external cloud platforms.
Sanitization: There is no evidence of sanitization or validation performed on the data ingested from the repository files before processing.

mj-deploy