Content
62%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.
The body delivers genuinely actionable, well-validated workflows with per-mode step lists, exact commands, and rigorous verification requirements. Its weaknesses are verbosity — the same scope, pass-budget, and reporting rules restated several times — and a set of references to non-existent bundle files (examples/, detector/) that undermines navigation.
Suggestions
Consolidate the pass-budget, verification-reporting, and protected-content rules, each of which is stated 3-4 times across the Editing contract, Modes, Marks pass, and Output format sections, into one authoritative statement referenced from the modes.
Fix or remove references to bundle paths that do not exist: examples/README.md, examples/<name>.json (cited as the --style config schema), and detector/validate.js (cited as the preservation validator).
Move the detailed Output-format policy and Tone-calibration material into a one-level-deep reference file, keeping a short per-mode checklist in SKILL.md to reduce the 317-line body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 317-line body restates the same policies many times: editing-pass budget rules appear in "Iterate to convergence," the marks pass, and three separate Output-format paragraphs ("A later repair that restores the original text still retains the passes actually used"); protected-content and scope rules are given in the Editing contract, again under Modes ("Follow the editing contract for protected material..."), and again under Edit-mode Verification. That is noticeably verbose with several padded sections (anchor 2), though it never explains concepts Claude already knows, so it is not a 1. | 2 / 5 |
Actionability | Guidance is largely concrete and executable: exact commands ("node scripts/normalize-quotes.js <rewritten-prose> --reference <original> --write", "node scripts/check-style.js <file> --config <path>"), explicit CLI flags, and a specified output structure (Issues found / Assessment; the four-item Verification checklist). It falls short of 5 because referenced tooling is not fully bundled — "examples/README.md", "examples/<name>.json", and "detector/validate.js" do not exist in the bundle — and no worked before/after rewrite example is given. | 4 / 5 |
Workflow Clarity | Each mode has a numbered job list (rewrite: audit → rewrite → summarize; edit: read → edit in place → verify), validation is explicit (re-read the file, compare final text against the source span-by-span, "Only proceed when validation passes"-style gates), feedback loops exist (repair within the remaining pass budget, re-verify), and the four-item Verification checklist covers the complex process. This matches the 5 anchor; a 4 would require missing checkpoints, and none are missing. | 5 / 5 |
Progressive Disclosure | The main reference is well signaled ("Before auditing or rewriting any text, read [references/patterns.md] in full" in a marked loading block) and the four scripts are referenced with exact invocations, but the body cites bundle paths that do not exist ("Schema and rationale: examples/README.md", "a bare name matching examples/<name>.json", "node detector/validate.js"), and ~150 lines of editing-contract and output-format policy are inlined in SKILL.md that could live one level deep. Structure is present with clear signaling, so above 2, but broken navigation and inline bulk keep it below 4. | 3 / 5 |
Total | 14 / 20 Passed |