apple-photos
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill requires the user to manually grant "Full Disk Access" to the terminal environment in macOS System Settings. While this is a documented prerequisite for the skill to access the Apple Photos SQLite database, it represents a significant increase in the agent's privilege level over the user's local filesystem.
- [COMMAND_EXECUTION]: The skill is built around the execution of multiple local shell scripts located in a
scripts/directory (e.g.,scripts/photos-export.sh,scripts/photos-search-content.sh). These scripts perform file operations and database queries on the host system. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the Apple Photos library, including filenames, person names, and machine-learning-generated content descriptions.
- Ingestion points: Photo metadata and search results are retrieved via scripts referenced in
SKILL.md(e.g.,photos-search-content.sh,photos-search-person.sh). - Boundary markers: None are defined to separate ingested data from agent instructions.
- Capability inventory: The skill can execute shell scripts and write files to the
/tmpdirectory as described in the "Workflow" section ofSKILL.md. - Sanitization: There is no mention of sanitization for the data retrieved from the Photos database before it is presented to the agent.
Audit Metadata