Browse And Evaluate
Search, preview and dry-run ai-agent-skills catalog entries before installing them
Test report
- Verdict
- Tested · Works
- Score
- Tested
- Jul 21, 2026
- Environment
- Claude Code 2.x (agent harness)
- Upstream re-checked
- Aug 10, 2026 · 9b1197f
Installed ai-agent-skills@4.3.2 from npm into a sandbox and ran every command the 58-line body prescribes. BASELINE (frontmatter only, no body): I dumped `--help` (~5KB into context), then `search database` (898 B) and `info database-design` (1877 B), with no dry-run before install. SKILL run: skipped --help entirely since the body supplies exact command shapes, `info --fields name,description,tags,collections,installCommands` returned 760 B vs the baseline's 1877 B (-59%), and `install database-design --dry-run -p` printed the exact write target and "bundled house copy from ..." before anything touched disk; the real install then wrote .agents/skills/database-design/SKILL.md. The biggest measured win is the --fields/--limit guardrail: bare `list` emits 68,770 bytes (~17k tokens) while `list --fields name --limit 3` emits 4 lines. Two honest dings: on `search` the prescribed --fields is a measured no-op (898 vs 896 bytes, those four fields are already the default), and the skill's frontmatter version 4.1.0 lags the 4.3.2 CLI. The Gotchas claim about preview stripping prompt injection is real — cli.js has PROMPT_INJECTION_PATTERNS (line 1574) and a `sanitized` boolean in the preview schema. Body references no bundled scripts or reference files, so nothing to 404; no curl|sh, base64 blobs, secret exfiltration or injection text anywhere in it.
Scored on four weighted criteria — install, triggering, output vs. baseline, docs. How scoring works
- Installs cleanly 5/5
- Triggers reliably 5/5
- Output vs. baseline 7/10
- Docs & honesty 4/5
What Browse And Evaluate does
Defines a five-step workflow for the ai-agent-skills CLI: search the catalog, pull skill details, preview the SKILL.md body, dry-run the install, then install. It enforces --fields, --limit and --dry-run flags so catalog output does not flood the context window and no skill is written to disk unreviewed. Triggers when you are browsing, comparing or evaluating skills in the ai-agent-skills catalog before installing.
How to install Browse And Evaluate
git clone --depth 1 https://github.com/MoizIbnYousaf/Ai-Agent-Skills.git /tmp/browse-and-evaluate-src
mkdir -p ~/.claude/skills
cp -R /tmp/browse-and-evaluate-src/skills/browse-and-evaluate ~/.claude/skills/browse-and-evaluate
# CLI instead: npx ai-agent-skills install browse-and-evaluate (global) or -p (project: .agents/skills/)
# Preview first: npx ai-agent-skills install browse-and-evaluate --dry-run
Skills live in ~/.claude/skills/ (global) or .claude/skills/
(per-project). Restart Claude Code after installing.
Commands — how to trigger Browse And Evaluate
-
/browse-and-evaluateSearch, preview and dry-run ai-agent-skills catalog entries before installing them
It also activates on plain-language prompts like these:
-
Find skills for database migration work -
Compare these two skills before I install one -
Dry-run installing this skill before committing
Frequently asked questions
- Is the Browse And Evaluate skill free?
- Yes. The skill itself is free from MoizIbnYousaf/Ai-Agent-Skills. SkillProof publishes the install command and an independent test verdict at no cost.
- Does Browse And Evaluate work with Claude Code?
- We tested it with Claude Code 2.x (agent harness) on Jul 21, 2026. Verdict: Tested · Works. Installed ai-agent-skills@4.3.2 from npm into a sandbox and ran every command the 58-line body prescribes. BASELINE (frontmatter only, no body): I dumped `--help` (~5KB into context), then `search database` (898 B) and `info database-design` (1877 B), with no dry-run before install. SKILL run: skipped --help entirely since the body supplies exact command shapes, `info --fields name,description,tags,collections,installCommands` returned 760 B vs the baseline's 1877 B (-59%), and `install database-design --dry-run -p` printed the exact write target and "bundled house copy from ..." before anything touched disk; the real install then wrote .agents/skills/database-design/SKILL.md. The biggest measured win is the --fields/--limit guardrail: bare `list` emits 68,770 bytes (~17k tokens) while `list --fields name --limit 3` emits 4 lines. Two honest dings: on `search` the prescribed --fields is a measured no-op (898 vs 896 bytes, those four fields are already the default), and the skill's frontmatter version 4.1.0 lags the 4.3.2 CLI. The Gotchas claim about preview stripping prompt injection is real — cli.js has PROMPT_INJECTION_PATTERNS (line 1574) and a `sanitized` boolean in the preview schema. Body references no bundled scripts or reference files, so nothing to 404; no curl|sh, base64 blobs, secret exfiltration or injection text anywhere in it.
- What is the Browse And Evaluate SkillProof Score?
- 8.4/10 — installs cleanly 5/5, triggers reliably 5/5, output vs. baseline 7/10, docs & honesty 4/5.
- How do I install Browse And Evaluate?
- Copy the install command from this page, run it in your terminal, and restart Claude Code. Skills live in ~/.claude/skills/ (global) or .claude/skills/ inside a project.
- Can I use Browse And Evaluate with Cursor, Copilot, Gemini CLI, Codex or other AI tools?
- The SKILL.md format is native to Claude (Claude Code, Desktop, claude.ai). The instructions inside adapt to other assistants: Cursor rules, GitHub Copilot instructions, Windsurf rules, Custom GPTs, AGENTS.md for OpenAI Codex, and GEMINI.md for Google Gemini CLI — our conversion guides cover each, and the free converter on the tools page does the wrapping for you.