Content
88%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 strong, highly actionable skill body: executable code throughout, explicit validation via the doctor with recovery loops, disciplined loop rules, and guardrails for risky actions. The two weaknesses are minor — some over-long sentences and slight install-guidance overlap cost conciseness, and the owned-engine reference chain adds a second hop of navigation.
Suggestions
Split the long Step 0 guard paragraph into shorter sentences or a small table (env value → guarded behavior → named error and response) to tighten conciseness; the run-on sentences about the connected-mode permission prompts and Stop handling are the densest in the file.
Flatten the owned-engine reference chain: add ladder.md, network.md, and frames-and-humans.md (with their one-line descriptions) directly to the 'Where the rest lives' table so every reference is one level deep from SKILL.md.
Deduplicate install guidance between Step 1 and the Rules section — the 'Never launch a headless browser because the attached one is missing' rule restates Step 1 content; keeping it in one place (or moving the detail to references/install.md) would save tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and imperative with no explanation of concepts Claude already knows, but a few passages could be trimmed: the long winding guard paragraph in Step 0 ("Under `connected` the app sees what the browser is doing, and before a click, Enter or script that sends, posts, pays, orders, subscribes, deletes or closes an account...") and install guidance that overlaps between Step 1 and the Rules section ("Never launch a headless browser because the attached one is missing" repeats Step 1's table content). This matches the anchor 'Efficient; minor instances of over-explanation that could be trimmed'; it is not 5 ('every token earns its place') because of those sentences, and clearly not 3 since there is no genuinely unnecessary explanation. | 4 / 5 |
Actionability | The guidance is fully executable and copy-paste ready: a working import snippet for `loadOmowright`, the command `node "<skill-root>/scripts/browser-doctor.mjs" --json` with a state-by-state table prescribing next actions, a complete try/finally session loop covering navigate, bskSnapshot, click, fill, press, waitForNavigation, screenshot, and stop, plus a concrete `requestHelp` example with real option names. This matches the anchor 'Fully executable; copy-paste ready code or commands; specific examples cover the common cases'. | 5 / 5 |
Workflow Clarity | The multi-step process is clearly sequenced with explicit validation and error-recovery loops: Step 0 (engine selection via `OMO_BROWSER_ENGINE` with a value table), Step 1 (load, run the doctor, and on missing pieces install then 're-run the doctor' — a validate-fix-retry loop), numbered loop rules including "Two identical failures mean change approach, not retry. A third identical attempt is a defect", and named error outcomes (`BrowserActionDeclinedError`, `BrowserUserStoppedError`) with prescribed responses. Destructive actions are explicitly gated (the app asks the user before anything that 'sends, posts, pays, orders, subscribes, deletes'), so the destructive-operation cap does not apply. This matches the anchor 'Clear sequence with explicit validation steps; feedback loops for error recovery; checklists for complex processes'. | 5 / 5 |
Progressive Disclosure | The body is a lean overview with a well-signaled 'Where the rest lives' table, and every referenced path exists in the bundle (commands.md, install.md, remote.md, owned-engine/README.md, recipes/1password.md, and the three scripts). However, the owned-engine branch is two levels deep: SKILL.md points to references/owned-engine/README.md, which is itself a routing index to ladder.md, network.md, and frames-and-humans.md. That minor deviation from the anchor's 'one-level-deep references' places this between the 4 anchor ('most content is appropriately placed; references mostly clear; minor organization gaps') and the 5 anchor; the README is a genuine routing summary with substantial inline content rather than a hollow pass-through, so it sits clearly above 3 but not fully at 5. | 4 / 5 |
Total | 18 / 20 Passed |