iw-research-quicklisted
Install: claude install-skill innovation-ways/iw-ai-core
# IW Research Quick
Fast inline research — no file, no ID, no workflow overhead.
**Topic**: $ARGUMENTS
---
## Workflow
1. **Get topic** — Extract from $ARGUMENTS or ask user a single clarifying question
2. **context7 (optional)** — If topic is a library/framework, try `mcp__context7__resolve-library-id` first for current official docs
3. **WebSearch** — Run 2–3 targeted queries to identify best sources
4. **WebFetch** — Fetch up to 4 pages maximum; extract targeted answers only
5. **Respond inline** — Direct answer (2–3 paragraphs) with:
- `[HIGH/MEDIUM/LOW]` confidence marker
- Inline source citations per factual claim
- `Sources` section at the end
6. **Upgrade nudge** — If topic warrants deeper investigation, suggest `/iw-research {topic}`
---
## When to Use context7
**Use context7 when** the topic is a **library or framework** and you want current official documentation:
```
mcp__context7__resolve-library-id <library-name>
mcp__context7__query-docs <library-id> <query>
```
**Do NOT use context7** for general topics (companies, concepts, market research, opinions, etc.).
If context7 is unavailable or returns no results, fall back to WebSearch + WebFetch only.
---
## Max 4 WebFetch Calls
**Hard limit: 4 WebFetch calls per invocation.**
Prioritize:
1. Official documentation (docs page, API reference)
2. Primary source (author blog, official announcement)
3. Credible secondary (well-known blog, reputable publication)
Skip: forums, Reddit threads, low-