agentic-productivity-setup
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and process untrusted local data, such as Git reflogs and AI agent session logs, which could potentially contain malicious instructions. * Ingestion points: Git reflogs and native agent registries (JSON/SQLite) as defined in the Collector Design. * Boundary markers: Instructions explicitly forbid storing or logging raw prompt content and mandate classification without retention. * Capability inventory: The system performs network POST requests to Discord, SQLite writes, and file system logging. * Sanitization: Implements one-way hashing of identifiers and strict local data aggregation.
- [DATA_EXFILTRATION]: The skill instructions define a data flow that transmits aggregate productivity metrics and locally rendered chart images to an external Discord webhook. This risk is mitigated by mandatory local-only rendering and the requirement to store the webhook URL securely in the macOS Keychain rather than in configuration files.
- [COMMAND_EXECUTION]: The skill manages credentials using the macOS security utility for Keychain access and installs background persistence using launchctl/LaunchAgents to handle daily reporting schedules.
Audit Metadata