photocraft-use
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches core engine binaries (CLI and desktop app) from the vendor's official GitHub repositories. All downloads occur over HTTPS and are verified against hardcoded SHA-256 digests before execution.
- [REMOTE_CODE_EXECUTION]: Executes the downloaded binaries to perform native image processing. The skill ensures safety by performing architectural checks (using lipo) and signature verification (using codesign) on the binaries before they are permitted to run.
- [COMMAND_EXECUTION]: Wraps the PhotoCraft CLI for orchestrated editing tasks. It uses a supervised execution model that monitors process identifiers and verifies the integrity of the inter-process communication via strictly validated JSON-RPC over stdio and local loopback TCP.
- [INDIRECT_PROMPT_INJECTION]: Ingests image assets and structured editing plans. It mitigates potential injection risks by strictly validating all command parameters against a pinned registry, enforcing path restrictions to authorized workspace roots, and employing a pixel-level guard to verify that non-target areas of the project remain untouched.
Audit Metadata