The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @vosjs/cli package via npm i -D @vosjs/cli. This is a vendor-owned resource associated with the skill author 'vosjs'.
[DATA_EXPOSURE]: The instructions direct the agent to access sensitive credentials in ~/.config/vos/credentials for authentication with the vendor's service. While this is part of the tool's intended functionality, accessing local credential files is a sensitive operation.
[COMMAND_EXECUTION]: The skill utilizes several CLI commands (vos fetch, vos pull, vos plan, vos digest, vos validate, vos frames, vos render, vos push) to perform its tasks, which involve subprocess execution and interaction with the file system.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources, creating a vulnerability surface.
Ingestion points: Reads recording.webm, cursor.json, meta.json, doc.json, digest.json, and project-specific .md recipe files.
Boundary markers: None explicitly defined for isolating instructions within the ingested JSON or Markdown data.
Capability inventory: File writing (edit doc.json), network operations (vos push), and shell command execution (npm install, vos CLI tools).
Sanitization: Uses vos validate to check markup validity, but no specific sanitization is mentioned for preventing instruction injection from metadata or recipe files.

vos-cut