task-forest

Pass

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses local Python scripts (scripts/task_forest.py and scripts/validate_task_forest_export.py) to manage task data. The analysis of these scripts shows they only perform file operations and local SQLite database updates within the workspace and a designated global registry folder in the user's home directory. Subprocess calls in the validation script are restricted to running the main skill logic using the system's Python interpreter.
  • [PROMPT_INJECTION]: The skill implements a strong mitigation against indirect prompt injection by using a proposal-based system. Any inferred changes to the task structure must be saved as a proposal which the user must explicitly confirm (via proposal-apply --yes) before the official task files are updated. This keeps the user in control of the task state.
  • [DATA_EXFILTRATION]: No network operations were found in the skill's code. All data is stored locally in the workspace under .agent-workbench/task-forest/ or in a local SQLite registry at ~/.agent-workbench/agent-workbench.sqlite3.
  • [SAFE]: The HTML export functionality correctly sanitizes user-provided data (like task titles and summaries) using HTML entity escaping before rendering the visualization, preventing potential cross-site scripting (XSS) risks when the user views the task graph in a browser.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 8, 2026, 11:00 AM
Security Audit — agent-trust-hub — task-forest