hermespet-macos-ai-companion
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches project source code from a GitHub repository and installs command-line utilities for AI services including anthropic-cli, @openai/codex-cli, and openclaw through official package managers.
- [COMMAND_EXECUTION]: Executes shell commands to build the application from source, including xcodebuild and custom scripts for creating distribution images (./scripts/create-dmg.sh).
- [PROMPT_INJECTION]: The memory management system provides a surface for indirect prompt injection.
- Ingestion points: MemoryManager.getDailySummary reads recorded user queries and file paths from the user_intents SQLite table.
- Boundary markers: Minimal; content is interpolated into a list format without structural delimiters or instruction-bypass warnings.
- Capability inventory: The generated summary is presented to the user and can influence subsequent AI engine interactions via the AIEngineManager, which controls multiple local and remote AI services.
- Sanitization: Uses a basic keyword blacklist and response truncation, but does not filter for instructional patterns in user-supplied query data.
Audit Metadata