Content
64%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured and highly actionable command reference for browser automation. Its main strength is the concrete, executable syntax for every command with clear examples. Its weaknesses are the monolithic length (too much reference material inline that should be in separate files) and the lack of error handling/validation guidance in workflows, which is important for inherently fragile browser automation tasks.
Suggestions
Move detailed command sections (mouse control, cookies/storage, network, tabs/frames, dialogs) into reference files and keep only the most common commands in SKILL.md
Add error recovery guidance to the core workflow — e.g., 'If snapshot returns no interactive elements, wait for page load: `agent-browser wait --load networkidle` then re-snapshot'
Add a troubleshooting section or validation checkpoints for common failure modes (element not found, navigation timeout, stale refs after DOM changes)
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is largely a comprehensive command reference, which is useful but quite long (~250+ lines). Much of it reads like a man page that could be split into reference files. However, it avoids explaining basic concepts and stays focused on commands. The sheer volume of commands listed inline (mouse control, cookies, storage, network, frames, dialogs) could be offloaded to reference docs. | 2 / 3 |
Actionability | Every command is concrete, copy-paste ready, and uses consistent syntax. The examples at the end (form submission, authentication) show complete executable workflows with realistic commands and expected outputs. No pseudocode or vague instructions. | 3 / 3 |
Workflow Clarity | The core workflow section provides a clear 4-step sequence, and the form submission example demonstrates a good flow. However, there are no explicit validation checkpoints or error recovery steps — e.g., what to do if a snapshot returns no interactive elements, if a click fails, or if navigation times out. For browser automation (which can be fragile), missing feedback loops and error handling guidance is a notable gap. | 2 / 3 |
Progressive Disclosure | The skill references deep-dive documentation and templates with a clear table structure, which is good. However, no bundle files are provided, so those references are unverifiable. More importantly, the SKILL.md itself is monolithic — the extensive command reference (mouse control, cookies, storage, network, frames, tabs, etc.) should be in reference files rather than inline, making the main file a wall of commands. | 2 / 3 |
Total | 9 / 12 Passed |