Prompt & Agent Workbench

Analyze prompts, track multi-agent session diagnostics, and explore automated hygiene optimization suggestions.

Ecosystem Prompt Quality Trend

Calculated aggregate score over last 30 days

Prompt Quality
86.4%
▲ 2.1% vs avg
Session Hygiene
92.1%
▲ 0.8% excellent
Code Review
88.0
▼ 1.5 pts linter warnings
Tool Mastery
95.4%
12 custom commands
Cost Burn
$184.50
Monthly Accum.
Runaway Loops
1
Reframe required

Session Explorer

Orchestrator Agent
Date Period
Active Project
Outcome Status
Prompt Analyzer Gate

Paste a coding prompt below to run local diagnostics before dispatching to model context.

0 characters

Analysis Summary

Prompt Quality Score: 88%
File Reference Check: PASSED
Context Scope Specificity: High
Coach Recommendations:
  • (Prompt looks solid. File references are correctly specified.)

Matched Sessions

session_8f2d9
Project: arche-site | 2026-05-26 00:04
Codex
Completed specification build for design-brief templates. All assets exported properly.
{ "id": "session_8f2d9", "agent": "Codex", "status": "success", "project": "arche-site", "prompt_quality": 0.88, "tokens": 14200, "cost": 0.071, "tool_calls": 3 }
session_7c9a1
Project: arche-site | 2026-05-25 23:51
agy (Self)
Refined marketing hero mesh rotation variables. Encountered tool loop warning: 9 read file retries.
{ "id": "session_7c9a1", "agent": "agy", "status": "warning", "project": "arche-site", "prompt_quality": 0.72, "tokens": 82400, "cost": 0.576, "tool_calls": 9 }
session_5d1f9
Project: sentinel | 2026-05-25 21:40
Hermes
Analyzed clinical intake records API and scaffolded mock database routers.
{ "id": "session_5d1f9", "agent": "Hermes", "status": "success", "project": "sentinel", "prompt_quality": 0.94, "tokens": 41900, "cost": 0.084, "tool_calls": 2 }

Optimization Suggestions Feed

Today, 00:09 — Quality Alert
Add file references to 3 prompts
Coach scanned your last 5 prompts for Codex. Three of them lacked explicit workspace file links, resulting in a 15% increase in research token overhead.
Yesterday, 19:12 — Guard Alert
Runaway Loop Detected (session_4e3a2)
Codex surpassed 12 tool operations (14 reached) without terminal state convergence. Engine reframe was auto-triggered. Recommend checking path exclusions.
May 24, 14:05 — Hygiene Suggestion
Session Exceeded 100K Tokens soft cap
Your active session with agy has reached 112,000 tokens. To maintain optimal speed and response accuracy, use the slash command `/new` to reset context.