The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection through the createSystemPrompt and setInputProps methods.
Ingestion points: Untrusted data enters the agent context via inputProps in the Helios constructor and the setInputProps method, as documented in SKILL.md.
Boundary markers: The documentation does not specify explicit boundary markers or delimiters for the inputProps data when it is interpolated into the prompt generated by createSystemPrompt.
Capability inventory: The skill allows network requests through registerStabilityCheck (e.g., fetch('/data.json')) and provides extensive control over composition state, playback, and DOM-synchronized animations.
Sanitization: The skill supports input validation via a HeliosSchema provided in the constructor, which can mitigate some risks if properly configured.

helios-core