photocraft-cli-resize

Pass

Audited by Gen Agent Trust Hub on Oct 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads and installs a fixed version of the PhotoCraft CLI and desktop application. It also requires the installation of the photocraft-harness skill via npx.
  • [COMMAND_EXECUTION]: The skill executes Python scripts from the harness directory (bootstrap.py, cli.py, workflow.py, commands.py) to manage installation, query commands, and run editing workflows. These executions use the -I -B flags for isolated execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect injection by processing untrusted images, assets, and JSON command plans.
  • Ingestion points: Processes external image assets (PNG, JPEG, etc.) and JSON-based command plans provided by the user or upstream tasks.
  • Boundary markers: Utilizes structured JSON schemas (craft-command-plan/v1) and explicit command registration to delimit instructions from data. Documentation notes the rejection of duplicate keys and malformed JSON.
  • Capability inventory: Performs subprocess calls to run Python scripts, interacts with the local file system for project saving/loading, and executes network operations for dependency installation.
  • Sanitization: Implements strict JSON validation, checks for non-finite values, and verifies file hashes (SHA-256) for project integrity and asset provenance.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 13, 2026, 01:38 PM