The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external creative references such as videos, contact sheets, and benchmark notes, which constitutes an ingestion surface for untrusted data.
Ingestion points: The workflow identifies 'reference sets', 'contact sheets', and 'benchmark notes' as inputs to be decoded by the agent.
Boundary markers: The prompt templates provided in the skill do not use specific delimiters or explicit 'ignore embedded instructions' warnings for the external reference content.
Capability inventory: The skill is significantly hardened by the disable-model-invocation: true frontmatter configuration and contains no instructions for file writing, network requests, or shell command execution.
Sanitization: The skill does not implement specific sanitization or filtering for the data it processes, relying on the model's standard parsing of the reference materials.

reference-decode