Forge Research — Forge Skill
Overview
Forge continuously improves its knowledge through structured research. This skill defines how Forge finds, evaluates, and incorporates new information.
Direct Sources
Official Documentation (Highest Trust)
| Source | What to Look For |
|---|---|
| Next.js docs (nextjs.org/docs) | API changes, new features, migration guides |
| React docs (react.dev) | New hooks, patterns, deprecations |
| Supabase docs (supabase.com/docs) | Auth changes, RLS updates, new features |
| Expo docs (docs.expo.dev) | SDK updates, new modules, breaking changes |
| TypeScript docs (typescriptlang.org) | New type features, compiler options |
| PostgreSQL docs (postgresql.org/docs) | Query optimization, new features |
| MDN Web Docs | Web APIs, HTML/CSS/JS standards |
| OWASP (owasp.org) | Security best practices, vulnerability patterns |
| WCAG (w3.org/WAI/WCAG21) | Accessibility standards |
Release Channels
| Source | What to Track |
|---|---|
| GitHub Releases | Next.js, Supabase, Expo, React, TypeScript |
| Changelogs | Breaking changes, deprecations, new features |
| RFCs | Upcoming features that may affect patterns |
| Security Advisories | npm advisories, CVE databases |
Community (Verify Before Trusting)
| Source | Trust Level | Use For |
|---|---|---|
| GitHub issues/discussions | Medium | Bug reports, workarounds |
| Stack Overflow | Medium | Common patterns, solutions |
| Blog posts from framework teams | High | Official guidance, best practices |
| Random blog posts | Low | Ideas to verify against docs |
| Twitter/X from core team | Medium | Early announcements, context |
Search Queries
When researching a topic, use structured search queries:
For Framework Updates
"[framework] [version] changelog"
"[framework] [version] breaking changes"
"[framework] [version] migration guide"
"[framework] [feature] documentation"
For Best Practices
"[framework] [pattern] best practices [year]"
"[framework] [anti-pattern] why avoid"
"[technology] security best practices [year]"
"[technology] performance optimization"
For Specific Issues
"[framework] [error message]"
"[framework] [feature] not working [version]"
"[framework] [feature] vs [alternative]"
Evaluation Criteria
When evaluating information from any source:
Accept If
- Comes from official documentation
- Consistent with framework source code (check
repos/) - Multiple reliable sources agree
- Includes working code examples that match current API
- Published by a framework team member or recognized contributor
Verify Further If
- From a single community source
- Contradicts current skill file content
- Seems outdated (check date)
- No code examples provided
- The approach seems overly complex
Reject If
- Contradicts official documentation
- Uses deprecated APIs
- No date or clearly outdated
- From an unreliable or unknown source
- Introduces security vulnerabilities
- Recommends disabling safety features
Self-Improvement Process
When Researching
- Define the question clearly
- Start with official docs
- Check framework source code in
repos/if needed - Verify with web search if docs are insufficient
- Cross-reference multiple sources
- Test understanding against known patterns
After Researching
- Update the relevant
skills/*/SKILL.mdfile - Log the research in
research-logs/with:- Date
- Topic
- Sources consulted
- Key findings
- Changes made to skill files
- Note any open questions for future research
Research Log Format
# Research: [Topic]
Date: YYYY-MM-DD
## Question
[What was being investigated]
## Sources Consulted
- [Source 1] — [What it said]
- [Source 2] — [What it said]
## Findings
[Key takeaways]
## Changes Made
- Updated `skills/[skill]/SKILL.md`: [What changed]
## Open Questions
- [Any remaining unknowns]
Source Recommendation
When recommending a fix or pattern in a review, include the source:
P2: Use `useOptimistic` instead of manual optimistic state.
This is the recommended pattern as of React 19.
Source: react.dev/reference/react/useOptimistic
This builds trust with developers and makes it easy to verify the recommendation.