agent-browser

Browse the web for any task — research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages. Use whenever a browser would be useful, not just when the user explicitly asks.

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

82%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A highly actionable, token-efficient CLI reference: fully executable commands, well-categorized sections, and worked examples that demonstrate the snapshot-ref-interact loop. The main improvements are adding error-recovery guidance for failed interactions and splitting the exhaustive command catalog into a one-level-deep reference file.

Suggestions

Add a short troubleshooting/error-recovery note for common failures (stale refs after DOM changes, element not found, timing-sensitive pages) to strengthen the feedback loop.

Move the exhaustive command catalog (Navigation, Get information, Cookies & Storage, JavaScript) into a references/COMMANDS.md and keep Quick start + Core workflow + examples in SKILL.md for progressive disclosure.

Trim the Quick start commands that are repeated verbatim in the Commands section, or merge the two to remove the duplication.

Dimension	Reasoning	Score
Conciseness	The body is a lean command reference with inline comments and no explanation of concepts Claude already knows; every command line carries real information. The only trim candidate is the minor duplication between the Quick start block and the Commands section (open/snapshot/click/fill/close appear in both), which keeps it just below the 'every token earns its place' anchor — well above level 3's 'unnecessary explanation'.	4 / 5
Actionability	Every snippet is a copy-paste-ready executable command, the full command catalog is concrete, and three worked examples (form submission, data extraction, auth with saved state) cover the common cases — matching the top anchor for fully executable, example-covered guidance.	5 / 5
Workflow Clarity	The Core workflow gives a clear 4-step sequence (navigate, snapshot, interact, re-snapshot) and the examples include verification checkpoints ('wait --load networkidle', 'snapshot -i # Check result'). It falls short of level 5 because there are no explicit error-recovery steps (e.g., what to do when a ref is stale or an element isn't found), though no destructive/batch cap applies.	4 / 5
Progressive Disclosure	Sections are well-organized with clear headers (Quick start, Commands by category, Examples) and no nested or buried references — but no bundle files exist and the entire CLI command catalog is inlined in SKILL.md rather than split into a references file, a minor organization gap matching the level-4 anchor 'most content is appropriately placed'.	4 / 5
	Total	17 / 20 Passed

Description

80%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong description: concrete action list, imperative voice consistent with the good examples, and an explicit 'Use when...' trigger clause. The main gaps are missing natural synonyms (scrape, log in, navigate) and a broad, generic trigger phrase instead of concrete trigger situations.

Suggestions

Add natural trigger synonyms users would actually say, e.g. 'scrape data', 'log in to a site', or 'navigate to a URL', to raise trigger-term coverage.

Make the 'when' clause more concrete with named trigger situations (e.g., 'Use when the user mentions screenshots, form filling, web scraping, or testing a web app') instead of the generic 'whenever a browser would be useful'.

Narrow 'for any task' to reduce overlap with research/data-extraction skills that may have their own dedicated skills.

Dimension	Reasoning	Score
Specificity	The description lists multiple specific concrete actions — 'research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages' — giving comprehensive coverage of the skill's capabilities, matching the top anchor.	5 / 5
Completeness	Both 'what' and 'when' are present ('Use whenever a browser would be useful, not just when the user explicitly asks'), but the 'when' clause is generic rather than the concrete trigger phrases of the level-5 anchor (e.g., 'when the user mentions screenshots or form filling'). Clearly above level 3, where 'when' is missing or only weakly implied.	4 / 5
Trigger Term Quality	Good natural keyword coverage ('browse the web', 'fill forms', 'take screenshots', 'extract data', 'web apps'), but common variations users would say — 'scrape', 'log in', 'navigate to a URL', 'open a page' — are missing. Not the level-5 anchor's 'comprehensive coverage including synonyms'; above level 3 because the terms present are ones users naturally say.	4 / 5
Distinctiveness Conflict Risk	Browser automation is a clear niche with distinct triggers, but 'Browse the web for any task' and 'whenever a browser would be useful' are broad and could pull in tasks overlapping with research, scraping, or testing skills — minor overlap risk with closely related skills, matching the level-4 anchor rather than level 5's 'minimal conflict risk'.	4 / 5
	Total	17 / 20 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: jbaruch/nanoclaw-telegram
Path: container/skills/agent-browser/SKILL.md
Commit: 7454673

Reviewed: 17 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.