<< All versions

Skill v1.0.1

currentAutomated scan100/100
cubetribe/claudecode_godmode-on/research

+4 new

──Details
PublishedOctober 23, 2026 at 06:23 PM
Content Hashsha256:6ec82ce4be827e01...
Git SHA73e4d0c382b2
Bump Typepatch
Compare with v1.0.0
──Files
Files (1 file, 3.5 KB)
SKILL.md3.5 KBactive
SKILL.md · 113 lines · 3.5 KB

version: "1.0.1" name: research description: "Research task workflow — @researcher agent for technology evaluation, best practices lookup, documentation discovery, with a non-enforced time budget guideline and memory guidelines" license: "Proprietary - (c) 2025-2026 Dennis Westermann. Free for private non-commercial use; redistribution/re-hosting prohibited. See LICENSE: github.com/cubetribe/ClaudeCode_GodMode-On"


Research Workflow

Command

"Research: [topic]"

Workflow

@researcher → report with sources

When to Use @researcher

ScenarioUse @researcher?
New library/framework neededYES
Best practices for unfamiliar domainYES
Security advisory checkYES
API documentation lookupYES
Well-known patterns (REST, MVC)NO
Bug fix in existing codeNO
Refactoring known codeNO

@researcher Configuration

  • Model: haiku (fast & cost-effective)
  • Tools: WebSearch, WebFetch, Read, Glob, memory MCP
  • Target budget: ~30 seconds per task (guideline, not an enforced limit — no

Timeout field on agents/researcher.md and no hook enforces it; see skills/cost-efficiency/ § Research Budget for the actual mechanism, which is scope discipline, not a clock)

Phase Budget (Guideline, Not Enforced)

These per-phase figures are a pacing target for the agent to self-report against, not a hard cutoff — nothing currently kills a task at these marks:

PhaseTargetIf running long
WebSearch~10sUse cached/partial results
WebFetch~8sSkip this source
Analysis~12sReport with available data
Total~30sProduce partial report

Report Format

markdown
# Research Report: [Topic]
## Summary
[1-2 sentence overview of findings]
## Findings
1.[Finding with source]
2.[Finding with source]
3.[Finding with source]
## Recommendation
[Clear recommendation with rationale]
## Sources
-[Source 1](URL) — [brief description]
-[Source 2](URL) — [brief description]
## Confidence
[HIGH/MEDIUM/LOW] — [reason]

Save to: reports/vX.Y.Z/sprint-NN/00-researcher-report.md

Graceful Degradation

If research fails or times out:

  1. Full Report — All sources found, all analysis complete
  2. Partial Report — Some sources found, partial analysis
  3. Failure Report — Research failed, structured error output

Partial report format includes completion_percentage and sources_status (FULL/PARTIAL/PENDING).

Memory Usage

After completing research:

  • STORE in memory: Key decisions, no-go technologies, verified sources
  • STORE when: High-confidence findings, deprecated tech discoveries
  • QUERY first: Check memory before starting new research (avoid duplicates)

Integration with Other Workflows

@researcher is OPTIONAL in Feature and API Change workflows:

Feature: (@researcher) → @architect → @builder → gates → @scribe
API Change: (@researcher) → @architect → @api-guardian → @builder → gates → @scribe

Use when the task involves unfamiliar technology. Skip when working with known patterns.


CC_GodMode — © 2025–2026 Dennis Westermann ([dennis-westermann.de](https://www.dennis-westermann.de)). Proprietary — not open source. Free for private, non-commercial use; redistribution or re-hosting outside GitHub is prohibited; attribution required. Official source: [github.com/cubetribe/ClaudeCode_GodMode-On](https://github.com/cubetribe/ClaudeCode_GodMode-On). See LICENSE.

← v1.0.0All versions