Content
82%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A dense, highly actionable reference: nearly every line is an executable command or a non-obvious pitfall, and it is organized into clear scenario-based sections. Its only real weaknesses are mild redundancy with Node's documented REPL basics and content (the CDP driver, profile recipes) that is large enough to justify splitting into bundle files.
Suggestions
Trim or condense the basic REPL command list (cont/next/step/bt) to only the non-obvious entries, keeping commands like sb('functionName') and the repl-mode locals trick that Claude is less likely to recall.
Move the CDP minimal driver and the heap/CPU profile capture recipes into scripts/ or references/ files, referenced one level deep from a short Programmatic CDP section, to keep SKILL.md as an overview.
Convert the Pitfalls list into explicit validate-then-proceed checkpoints inside the debugging flows (e.g., verify /json/list shows the target before attaching) to strengthen feedback loops.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — bullet commands, a compact CDP driver, and high-value pitfalls with no padded prose. Minor trimmable content remains, such as the basic REPL command list ("cont, next, step, out, pause") and "bt"/"list(8)" which are standard documented node inspect commands Claude largely already knows. | 4 / 5 |
Actionability | Everything is copy-paste ready: exact invocations ("node inspect path/to/script.js", "node --inspect-brk --import tsx ...", "kill -SIGUSR1 <pid>"), a curl/jq inspection command, and a complete executable CDP driver covering breakpoint, evaluate, and resume flows. | 5 / 5 |
Workflow Clarity | Scenarios are clearly sequenced: quick-start entry points, REPL controls with an explicit safety checkpoint ("cont before quitting if the process should continue; otherwise kill"), a staged CDP path (install tooling → driver → profiles), and pitfall-recovery guidance ("If a breakpoint misses, confirm file path, source map behavior..."). It is a reference-style skill rather than a multi-step pipeline, so feedback loops are present but informal rather than explicit validate-then-proceed steps. | 4 / 5 |
Progressive Disclosure | No bundle files exist and the body references none; sections are well-organized (Quick start, Debugger REPL, OpenClaw tips, Programmatic CDP, Profiles, Pitfalls) and easy to navigate. At ~70 lines it exceeds the under-50-line exception for a clean 5, and the ~25-line CDP driver plus profile recipes are arguably large enough to live in a scripts/ or references/ file, matching anchor 4's "minor organization gaps". | 4 / 5 |
Total | 17 / 20 Passed |