task-forest
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The validation script
scripts/validate_task_forest_export.pyusessubprocess.runto execute the skill's own CLI for automated testing. This is a standard development practice for regression testing and does not involve executing arbitrary or untrusted external commands. - [INDIRECT_PROMPT_INJECTION]: The skill processes task definitions and proposals from local JSON files. While this creates a surface for indirect prompt injection via manipulated task data, the risk is effectively mitigated by the skill's workflow, which requires explicit user review and confirmation before any changes are applied to the canonical task graph.
- [DATA_EXFILTRATION]: The skill provides an opt-in feature to track workspace health in a local SQLite database at
~/.agent-workbench/agent-workbench.sqlite3. This operation is confined to the local system and is used solely for project discovery and status reporting, with no network connectivity or data transmission to external servers observed.
Audit Metadata