The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code and configuration data from external programs fetched from the gallery. 1. Ingestion points: config.json files fetched via vos fetch from https://vos.so (documented in SKILL.md). 2. Boundary markers: The instructions do not specify explicit delimiters to isolate fetched JavaScript strings from the agent's internal instruction context. 3. Capability inventory: The skill utilizes the @vosjs/cli to compile code (vos check), render stills (vos still), and push updates to the platform (vos push). 4. Sanitization: The vos platform performs server-side compilation and schema validation, while the skill recommends local checking and rendering to catch runtime errors.
[DYNAMIC_EXECUTION]: The skill involves editing and executing JavaScript strings locally to validate program behavior. 1. Execution Method: The vos still command executes the compiled JavaScript module to render a preview frame as part of the development loop. 2. Context: This behavior is the primary purpose of the skill and is performed using the official vendor CLI tools.

vos-remix