The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes local workspace content to provide animation audits and suggestions, creating a potential surface for indirect prompt injection. This is restricted to the skill's primary function of code analysis and lacks dangerous capability chains.
Ingestion points: Workspace files (CSS, JSX, TSX) are read by the agent via the transitions review and transitions refine commands.
Boundary markers: Not explicitly specified in the skill instructions.
Capability inventory: File-write operations are performed during the transitions apply phase to install snippets.
Sanitization: No explicit sanitization of workspace content is described.
[SAFE]: The skill operates entirely on locally provided markdown files containing CSS and JavaScript snippets. It does not perform any remote network operations, data exfiltration, or external script downloads.

transitions-dev