Loading ToolBench
Run and inspect Codex workflow-as-code jobs. Prefer workflow_run for release review fanout, workflow_status for concise progress, and workflow_events only when raw event history is needed.
Server defines 11 tools with complete input schemas and descriptions. Naming follows verb_noun convention (workflow_run, workflow_status, etc.). Descriptions are present but brief (10-50 chars), below the 50-200 char baseline for LLM optimization. All parameters have type definitions and descriptions. Output schemas are not documented in the provided source. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications (WRITE, IRREVERSIBLE, READ_ONLY). Parameter descriptions lack constraint details (e.g., enum values not explained in text, format expectations missing).
Get raw event history for a workflow run.
List built-in and saved workflows.
Open a terminal dashboard for monitoring a workflow run.
Pause a running workflow.
Restart a failed agent in a workflow.
Resume a paused workflow.
Start a Codex workflow run.
Get concise workflow run status and progress.
Tool descriptions are too brief (10-50 chars vs. 50-200 char baseline). LLMs cannot determine when to select tools or understand prerequisites. E.g., 'Start a Codex workflow run' lacks context on when to use it vs. workflow_resume, what args are required, or what the return value contains.
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., runId from workflow_run). Pattern requires documenting return types for all tools.
No tool annotations despite clear risk classifications. workflow_stop and workflow_stop_agent are IRREVERSIBLE but lack destructiveHint. workflow_status and workflow_events are READ_ONLY but lack readOnlyHint. Agents cannot reason about safety without annotations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Stop a running workflow.
Stop a specific agent in a running workflow.
Validate a workflow definition file.
Parameter descriptions lack constraint details. Enum values (adapter: auto|simulate|exec|sdk, storageScope: codex-home|project) are defined in schema but not explained in description text. LLMs cannot read JSON Schema, they rely on description text to understand valid options.
No error handling or recovery guidance. Tools do not document what errors can occur, whether they are retryable, or what the LLM should do next. E.g., workflow_run could fail if workflowPath is invalid, but no guidance is provided.