Loading ToolBench
Governance layer for AI agent teams with audit trail, process management, dependency analysis, and decision support
Server has 21 tools with basic schemas and descriptions, but significant gaps limit production readiness. Most tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization) and partial input schemas. Critical issues: many parameter descriptions are minimal or missing; no output schemas documented; error handling is absent; no tool annotations (readOnlyHint/destructiveHint); composition issues (org_read_artifacts description says 'audit log' but tool name suggests artifacts); naming inconsistencies (org_read_* vs org_update_* vs org_create_* lack clear verb-noun patterns for some operations). READ_ONLY tools are well-separated from WRITE tools, but descriptions don't explain WHEN to use each tool or what downstream tools expect.
Calculate semantic energy E(x) (quadratic convergence)
Calculate Hamiltonian priority H(s)
Create a new process (P.0)
Register a new tension
Add a dependency between processes
Full dependency analysis (FDM) - cycles, topological sort, parallel groups, impact scores
Remove a dependency between processes
Show dependencies for a specific process
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., process_id, agent_id) from responses. Violates pattern:tool and pattern:response-shaper.
Many tools lack input schemas entirely (org_read_state, org_read_tensions, org_read_attractors, org_read_agents, org_read_health, org_verify_chain). Per HARD SCORING RULE, schema score must be 0 for these tools.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | <=2025-11-25 | v2 |
Pull latest, optionally commit+push changes
Append to artifacts.jsonl (auto-calculates E(x) for V-steps)
Read registered agents
Read recent entries from audit log
Read strategic attractors/goals
Read convergence history for a process
Read health metrics
Read a specific process and its artifacts
Read current process states
Read open tensions
Mark a tension as resolved
Update a process state
Verify hash-chain integrity of audit log
Descriptions are too short (10-55 chars vs. baseline 50-200 chars for LLM optimization). org_read_state, org_read_tensions, org_read_attractors, org_read_agents, org_read_health all have 30-char descriptions that lack context for tool selection. Per HARD SCORING RULE, descriptions under 20 chars score 0-20.
No error handling guidance. Tools do not return actionable error messages (e.g., 'process_id not found, try org_read_state() to list valid processes'). Violates pattern:recovery-guide and pattern:error-classification.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot distinguish safe reads from destructive writes without explicit hints. org_update_state, org_create_process, org_log_artifact, org_create_tension, org_resolve_tension, org_git_sync, org_deps_add, org_deps_remove should be marked destructiveHint=true.