h3-cinematic-director
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown and local reference files. It does not include any scripts, executables, or commands that interact with the host system, network, or sensitive environment variables.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, specifically scripts and production briefs provided by users.
- Ingestion points: Data enters the skill's logic through user-provided scripts, shot briefs, and storyboards passed as parameters to invocation commands in
SKILL.md. - Boundary markers: The skill contains an explicit 'Information Boundary' section in
SKILL.mdthat restricts the agent to using only supplied assets and forbids the retrieval of private content from other projects. - Capability inventory: The skill is limited to text and prompt generation; no capabilities for file-writing, network communication, or system command execution are present in the skill files.
- Sanitization: The instructions include a 'Non-Negotiable Quality Bar' and a 'Continuity Audit' workflow in
references/continuity-audit.mdthat force the agent to validate all outputs against canonical assets, reducing the risk of the model following instructions embedded within the processed data.
Audit Metadata