filmcraft-interview-edit

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation and metadata contain references to external resources and download links for supporting tools.
  • Evidence: references/craft-command-evidence.json lists a GitHub release URL for the FilmCraft CLI (github.com/storytold/filmcraft).
  • Evidence: references/market-adaptations.json lists a reference source with a download link from skillhub.cn.
  • Note: These are informational references for the tool and research sources, not automated execution instructions.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as an interface for processing external interview media, which constitutes a potential attack surface for indirect prompt injection via media metadata or transcript content.
  • Ingestion points: Media materials are imported using open_file as described in SKILL.md.
  • Boundary markers: The workflow includes explicit steps to inspect and verify sequences using inspect before final delivery.
  • Capability inventory: The skill utilizes command_run for executing editing operations and save_project for persistence.
  • Sanitization: Instructions focus on manual and semantic verification of output content rather than automated technical sanitization.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 10:16 AM