The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches version information and updated instructions from raw.githubusercontent.com/fredm00n/framerlabs. This is used to ensure the agent has the latest techniques for Framer development.
[PERSISTENCE]: The skill contains instructions for the agent to automatically update its own local files. If the auto-update flag is present in the project configuration, the agent will clone the fredm00n/framerlabs repository and replace the installed skill folder with the new version.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic imports to load external libraries such as hls.js from esm.sh at runtime. Additionally, it provides patterns for accessing React internal properties (__reactFiber and __reactProps) to trigger interaction handlers that are not exposed via standard DOM events.
[COMMAND_EXECUTION]: The instructions include steps for the agent to run commands like git clone for updates and contributions, and mentions using npx skills add to manually update the skill.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from the Framer CMS and URL parameters, creating an ingestion surface for untrusted data. While it includes some data processing logic, there are no explicit boundary markers to prevent potential prompt injection from these sources. Ingestion points: props.text (CMS content), props.children (element tree), and window.location.search (URL params). Boundary markers: Not explicitly present in instructions for data processing. Capability inventory: Includes network requests via fetch, URL manipulation via history.replaceState, and DOM modification via MutationObserver. Sanitization: Instructions are provided to sanitize project-specific data (keys, tokens) before contributing findings back to the repository, but no specific sanitization is mentioned for runtime data ingestion.

framer-code-components-overrides