MCP Server for Plan Repair Dynamic Re-Anchoring Skill. Repairs broken execution plans when workflow steps fail by calculating affected subgraphs, preserving valid downstream branches, and splicing in alternative recovery actions.
Two tools with basic schemas and minimal descriptions. register_step and repair_plan both lack parameter descriptions, critical for LLM understanding. Descriptions are present but generic (under 100 chars). No output schema documented. No error handling guidance. No tool annotations. Schemas are present but parameters lack descriptions, violating the 100% baseline for A+ tools. This server falls into the 'poor' category typical of community implementations.
Register workflow DAG step with dependencies and fallback actions
Dynamically repair broken plan when a step fails
Parameter descriptions missing entirely. step_name, prerequisites, fallback_actions in register_step have no descriptions. LLMs cannot infer parameter meaning from names alone.
No output schema documented. Tools return JSON but LLMs have no formal specification of response structure, fields, or types. Agents cannot plan downstream calls reliably.
Tool descriptions are generic and under 100 chars. 'Register workflow DAG step with dependencies and fallback actions' lacks WHEN to use, prerequisites, or side effects. No guidance on state mutation.
No error handling or recovery guidance. If repair_plan returns repair_possible=false, LLM has no actionable next step. No error classification (retryable vs fatal).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). register_step mutates state but LLM cannot infer this from the definition. repair_plan is read-only but unmarked.