Content
87%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
Well-organized and actionable content that delegates detail appropriately to existing reference files. The main gap is the absence of explicit validation/verification checkpoints in the snapshot and root-cause workflows.
Suggestions
Add an explicit validation step after capturing snapshots (e.g., confirm the leak signal appears in the target-vs-baseline comparison before moving to root-cause identification).
Introduce a feedback loop around the 10x-interaction capture step: verify the leak reproduces/amplifies before taking the final snapshot, so a false negative is caught early.
For the manual fallback, state how to interpret the script's output and what to do if no leak type is flagged (retry with a different baseline or longer interaction window).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and directive, assuming Claude's competence (no explanation of what a memory leak or heapsnapshot is); the only mild restatement is the opening sentence and brief guardrail rationale, both earning their place. | 3 / 3 |
Actionability | Provides a fully executable fallback command (node ... compare_snapshots.js <baseline> <target>) plus concrete named MCP tools (click, navigate_page, take_heapsnapshot) and quantified steps ('Repeat the same user interactions 10 times'). | 3 / 3 |
Workflow Clarity | Four workflows are clearly sequenced, but the snapshot-capture and leak-identification flow lacks explicit validation checkpoints (e.g., confirm a leak signal before root-causing), and the batch 'repeat 10 times' step has no verify-feedback loop, capping this at 2. | 2 / 3 |
Progressive Disclosure | The body is an overview that splits detail into three clearly signaled, one-level-deep references (memlab.md, common-leaks.md, compare_snapshots.js), all of which exist on disk, with no nested references. | 3 / 3 |
Total | 11 / 12 Passed |