photocraft-cli-selection

Pass

Audited by Gen Agent Trust Hub on Oct 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses a bootstrap script to download and install a pinned version of the PhotoCraft CLI to a local user data directory. This is consistent with the skill's primary purpose as a CLI wrapper.
  • [COMMAND_EXECUTION]: Multiple Python scripts (e.g., bootstrap.py, cli.py, workflow.py, commands.py) are executed to handle installation, command routing, and workflow orchestration. These scripts are sourced from the companion photocraft-harness skill provided by the same vendor.
  • [REMOTE_CODE_EXECUTION]: The skill instructs the user/agent to install dependencies using npx skills add full-aigc-skills/photocraft-skills. It then executes local Python scripts to bootstrap the environment and run binary components.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user assets and image projects. It implements a mandatory evidence chain to mitigate risks:
  • Ingestion points: Assets registered via --asset and project files loaded through doc_open (e.g., in workflow.py).
  • Boundary markers: The skill supports explicit root isolation via --automation-read-root and --automation-write-root parameters.
  • Capability inventory: The skill performs file writes (doc_save), exports (doc_export), and executes the PhotoCraft binary through its gateway scripts.
  • Sanitization: All project modifications are guarded by expectedProjectSha256 and expectedManifestSha256 hash verifications to ensure that only authorized versions are modified.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 13, 2026, 01:36 PM