photocraft-cli-filters
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses Python wrapper scripts to interface with the PhotoCraft CLI. These commands are restricted to a defined set of creative operations and include state checks to ensure commands are only executed when the project context is valid.
- [EXTERNAL_DOWNLOADS]: The skill downloads and installs the PhotoCraft CLI and associated desktop components. Security is maintained through the use of fixed official URLs, architecture-specific binary pinning, and integrity verification of the entire installation tree against locked digests.
- [INDIRECT_PROMPT_INJECTION]: The skill processes complex image project files and JSON-based operation plans, which creates a potential surface for indirect prompt injection. To mitigate this, the skill implements robust boundary markers, such as mandatory integrity manifests for every delivery, strict JSON schema validation, and read-only verification tools that inspect project structures before editing.
Audit Metadata