The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of .drx grade files, .drp project files, and rendered image frames, which represent an ingestion surface for untrusted content.
Ingestion points: As described in SKILL.md, the skill reads external metadata and frames via tools such as scope_read, match_to_reference, and media_analysis.
Boundary markers: The instructions specify the use of validation routines like verify_grade and check_version_support to ensure expected behavior.
Capability inventory: The skill has the capability to modify the DaVinci Resolve node graph, write to local project databases, and generate .drx files across multiple scripts and servers.
Sanitization: The skill utilizes deterministic numpy-based assessments and auto-canonicalization of hue-axis curves to validate and normalize processed data.
[COMMAND_EXECUTION]: The skill interacts with the host environment through specialized toolsets for live and offline grading.
Evidence: The skill routes operations to a live Python server (davinci-resolve) for timeline manipulation and an offline Node server (davinci-resolve-advanced) for file-based grade computation as detailed in SKILL.md.

resolve-color