paw-cra-agent-designer
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: Uses
curlcommands to interact with the fal.ai image generation API, passing parameters for prompts, aspect ratios, and model selection.- [EXTERNAL_DOWNLOADS]: Fetches generated image assets from the fal.ai platform (fal.run) and saves them to the local workspace in the.pawbytes/creative-suites/directory.- [DATA_EXFILTRATION]: Retrieves thefal_keyfrom local configuration files (.pawbytes/config/config.user.yaml) to authenticate image generation requests to the external fal.ai service.- [PROMPT_INJECTION]: The skill includes a research capability that ingests information from the web to update its knowledge base, creating a surface for potential indirect prompt injection. - Ingestion points: Web search results processed in
references/research-capability.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified.
- Capability inventory: Shell execution via
curland file-write access across the.pawbytes/directory. - Sanitization: No sanitization or validation of external researched content is mentioned before it is stored or used.
Audit Metadata