The Agent Skills Directory

[COMMAND_EXECUTION]: Orchestrates the agent or cursor-agent CLI through shell wrappers (bin/composer-run.sh, bin/cursor-agent-doctor.sh) to automate AI-driven development tasks.
[DATA_EXFILTRATION]: Accesses environment variables and local .env files to retrieve CURSOR_API_KEY. It implements redaction logic in bin/cursor-agent-lib.sh to prevent these credentials from appearing in command output or logs.
[PROMPT_INJECTION]: The skill processes untrusted external data which serves as a surface for indirect prompt injection.
Ingestion points: Ingests git diffs and PR patches in review.md and natural language implementation briefs in generate.md.
Boundary markers: Lacks explicit delimiters or instructions to ignore malicious commands embedded in the processed diffs or briefs.
Capability inventory: Executes shell commands, performs file system writes, and manages git repositories via the wrapped agent.
Sanitization: Does not perform sanitization or validation of the content of ingested diffs or briefs before passing them to the AI agent.
[EXTERNAL_DOWNLOADS]: Recommends the installation of the Cursor Agent CLI using a script from the well-known service cursor.com (curl https://cursor.com/install -fsS | bash).

composer