Content
100%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a lean, highly actionable overview: concrete tooling directives, a fully sequenced workflow with validation gates and a repair loop, and a well-signaled one-level reference index. It respects Claude's intelligence and earns its tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body assumes Claude's competence — it never explains what MPS, an AST, or closures are — and every line is operational guidance, so each token earns its place. | 3 / 3 |
Actionability | Concrete tool names (e.g. `mps_mcp_parse_java_and_insert`, `mps_mcp_update_node`), exact concept/role names, and specific directives (e.g. 'ConstructorDeclaration.returnType MUST be a VoidType node') give fully executable guidance with no pseudocode. | 3 / 3 |
Workflow Clarity | The 8-step Common Workflow is clearly sequenced with explicit validation checkpoints ('Validate in three gates': dryRun, check_root_node_problems, alter_nodes MAKE) and a 'Repair' feedback loop via FIX_REFERENCES, matching the top anchor. | 3 / 3 |
Progressive Disclosure | The Reference Index signals seven one-level-deep, real files (all present under references/) with a clear 'Open … when you …' navigation cue for each, keeping the body an overview pointing to detail. | 3 / 3 |
Total | 12 / 12 Passed |