alterlab-cdm-vfx-pipeline
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze potentially untrusted data from local project files and web search results, creating an attack surface where embedded instructions in those sources could override the agent's behavior.
- Ingestion points: The agent reads existing project files (scripts, storyboards, notes) and performs web searches for technical documentation and tutorials.
- Boundary markers: The instructions lack delimiters or warnings to ignore instructions within analyzed data, which could lead the agent to obey commands found in the input files.
- Capability inventory: The skill can read files, write structured markdown files to the local directory, and perform outgoing web searches.
- Sanitization: No sanitization of external data is specified before it is used to generate deliverables or used in search queries.
- [DATA_EXFILTRATION]: The combination of reading local project files and performing outgoing web searches creates a potential data exposure surface. While no explicit exfiltration logic is present, the agent could be manipulated via prompt injection to include sensitive content from analyzed files in its search queries.
Audit Metadata