The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including Figma design frames and Pull Request (PR) content, which could theoretically contain malicious instructions.
Ingestion points: Reads Figma URLs (/cursor-build-figma) and PR data (/cursor-dyl-review, /cursor-dyl-ready-pr) as specified in SKILL.md and playbooks.
Boundary markers: The skill includes explicit warnings in references/requirements.md and SKILL.md stating that tools must be resolved against the host, schemas must be verified, and system/developer instructions retain priority.
Capability inventory: Includes file system modification (code editing), command execution (gh pr, origin pr), and network interactions via Figma MCP tools.
Sanitization: Relies on native agent permissions and explicit operator authorization requirements for sensitive actions like merging.
[COMMAND_EXECUTION]: The skill utilizes standard command-line interface (CLI) tools for repository management and UI automation.
Evidence: References to gh pr (GitHub CLI) and origin pr are found in playbooks/opening-a-pr.md. Usage of cursor-control-ui and cursor-control-cli for environment interaction is documented in playbooks/bug-fix.md.
[EXTERNAL_DOWNLOADS]: The skill identifies external dependencies and source origins for informational purposes.
Evidence: MIRROR.md references the source repository at https://github.com/cursor/plugins.git. references/requirements.md lists external capabilities such as cursor-poteto-mode and cursor-team-kit which must be resolved locally.
Context: The skill explicitly forbids the automatic installation of any external tools or plugins, directing the agent to report missing capabilities rather than attempting to acquire them.

cursor-dyl-mode