Content
85%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 highly actionable, well-sequenced skill body with executable examples and strong workflow/validation guidance. The main defect is progressive disclosure: dead reference and template links, plus a large inline command reference that should be split out, undercut an otherwise clean one-level-deep structure.
Suggestions
Create references/commands.md and references/profiling.md (or remove those rows from the Deep-Dive Documentation table) so every linked file resolves.
Ship the templates/ directory referenced in 'Ready-to-Use Templates' or delete that section — three dead links currently point at nonexistent files.
Move the bulk 'Essential Commands' block into references/commands.md, keeping only the handful of commands needed for the Core Workflow inline in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and command-focused with almost no explanation of concepts Claude already knows, e.g. the Essential Commands section is pure executable reference. Not 5 because the ~55-line inline Essential Commands block duplicates material that belongs in the (linked) full command reference, and the "Why this matters" paragraph under eval re-explains shell quoting behavior at length. | 4 / 5 |
Actionability | Nearly every section is copy-paste-ready with real flags and worked end-to-end examples — e.g. "agent-browser open https://example.com && agent-browser wait --load networkidle && agent-browser snapshot -i", the eval "--stdin <<'EVALEOF'" heredoc pattern, and concrete guidance on when to chain vs. run separately. Common cases are covered with specific commands and expected output. | 5 / 5 |
Workflow Clarity | The Core Workflow (Navigate → Snapshot → Interact → Re-snapshot) is stated up front and reinforced by a dedicated "Ref Lifecycle (Important)" section, the diff workflow ("snapshot -> action -> diff") provides an explicit verification loop, and timeout/slow-page plus session-cleanup guidance closes the sequence. Re-snapshotting after DOM changes is the explicit validation checkpoint for this skill's fragile operation (stale refs). | 5 / 5 |
Progressive Disclosure | Structure and signaling are good — a Deep-Dive Documentation table with when-to-use guidance, and five real reference files — but two of the seven linked references (references/commands.md, references/profiling.md) do not exist, and all three templates/*.sh links point to a nonexistent templates/ directory. Additionally, the bulk Essential Commands reference is inlined in SKILL.md instead of living in the separate commands file it links to. This matches the anchor of structure present but content that should be separate is inline, compounded by dead links. | 3 / 5 |
Total | 17 / 20 Passed |