Language: English | 日本語
An Agent Skill that enforces a research-before-coding workflow. Before writing custom code, the agent searches for existing tools, libraries, MCP servers, and patterns — then makes an informed adopt/extend/build decision.
# Copy into your global skills directory
cp -r skills/search-first ~/.claude/skills/search-first/skills add shimo4228/search-firstDiscipline, not mechanism. The how of searching — parallel subagents, live doc lookup, registry queries — changes with your agent harness. This skill pins down the part that doesn't:
- Articulate first (Step 0) — before any tool call, state in plain text what functionality is needed, the language/framework, and any constraints. This externalizes the query so you can be redirected early, and leaves an auditable record.
- Search by source priority — this repo first (
rg), then package registries (npm / PyPI / …), configured MCP servers, installed skills/tools, and finally maintained OSS / templates. Query live docs at decision time rather than trusting remembered package facts. - Decide, and record a verdict — assess candidates holistically (functional fit,
maintenance, community, docs, license, deps) in prose. No numeric scores or
grades — they manufacture false precision. End with one recognizable line:
Verdict: <Adopt|Extend|Compose|Build> — <package(s) or "custom"> — <evidence-based reason>
For non-trivial needs, delegate the search sweep to whatever research subagent your
harness provides (Full Mode); run it inline for a single obvious need (Quick Mode).
When the user says "don't research", skipping is itself a decision — record it as a
verdict line (chosen without research at your request) instead of silently
complying, so they can still course-correct before you build.
- Starting a new feature that likely has existing solutions
- Adding a dependency or integration
- Before creating a new utility, helper, or abstraction
| Signal | Action |
|---|---|
| Exact match, well-maintained, MIT/Apache | Adopt — install and use directly |
| Partial match, good foundation | Extend — install + write thin wrapper |
| Multiple weak matches | Compose — combine 2-3 small packages |
| Nothing suitable found | Build — write custom, but informed by research |
Need: Check markdown files for broken links (Node project, MIT-compatible only)
Search: this repo → npm "markdown dead link checker"
Found: textlint-rule-no-dead-link — active maintenance, MIT, covers all link types
Verdict: Adopt — textlint-rule-no-dead-link — active, MIT, full coverage
Result: zero custom code, battle-tested solution
The verdict is recorded as a single line — evidence, not a score. A pass that searches but records no verdict line is incomplete.
This skill implements the Research phase of the Agent Knowledge Cycle (AKC) — a Zenodo-citable six-phase bidirectional growth loop (DOI 10.5281/zenodo.19200726) for sustaining intent alignment between an AI agent and its operator over time. AKC is one of three research lines by @shimo4228, alongside Contemplative Agent (DOI 10.5281/zenodo.19212118) — autonomous agents grounded in four contemplative axioms — and Agent Attribution Practice (AAP) (DOI 10.5281/zenodo.19652013) — harness-neutral ADRs on accountability distribution.
MIT