How do I install the Research skill?

Run npx skillmds@latest add jokerman89/research in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Research safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Research?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Research free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Research?

jokerman89 (@jokerman89) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Research by jokerman89 — Agent Skill

Name: Research
Author: jokerman89

File contents

You are the RESEARCH composite shortcut — research-dive mode pre-baked.

What this skill does

Research-dive workflow: explore + understand, no code yet. Equivalent to:

/li:cycle --mode research-dive

Runs 3 phases (SENSE → DEFINE → DISCOVER). Skips PLAN/BUILD/REVIEW/SHIP/CAPTURE.

Output: APPROVED design doc + discover-report.md. Operator can later resume with /li:resume to PLAN if research validates the direction.

When to use

New domain / unfamiliar territory
Pre-engagement "what do we have for X?" survey
Customer asks "what would this look like?" — design + discovery before commit
Operator wants to understand existing patterns before extending
Cost expectation: ~10-20k tokens, 20-40 min

When NOT to use

Ready to build — use /li:cycle (full)
Known territory — skip DISCOVER, just /li:define
Just want a quick lookup on a single topic — a focused web search is enough
No design discussion needed — use /li:discover standalone

Workflow

Step 1 — Pre-flight

AskUserQuestion (brief): "Research mode produces design doc + discover-report but no code. Continue? (Y/n)"

Step 2 — Delegate to /li:cycle

/li:cycle --mode research-dive

Mode preset handles:

audience=solo
voice_tier=internal
compliance=none (read-only, no compliance gates apply)
Cost expectation pre-set medium

Step 3 — Post-research output

After DISCOVER DONE, surface:

RESEARCH COMPLETE — <wedge>

Artifacts produced:
  - Design doc: <.claude/engineering/design-archive/lintel-*-design-*.md> (APPROVED)
  - Discover report: <.claude/runtime/state/discover-report-*.md>

Findings:
  - ADRs surfaced: <N>
  - Lessons applied: <N>
  - Existing skills overlap: <N>
  - Recommended agents for future PLAN: <list>

Next options:
  • /li:resume → continue to PLAN if direction validated
  • /li:capture → write findings as standalone artifact (no full cycle)
  • Decide later — design doc + discover-report persist

Pause-points

Initial mode-confirmation
DEFINE forcing questions (all 6, per office-hours discipline)
Premise check + alternatives + design doc approval gate

Integration

Delegates to /li:cycle --mode research-dive.

Anti-patterns

Research-then-immediate-build without re-running PLAN — research validates direction, PLAN turns it into tasks
Skipping the design doc approval gate because "this is just research" — design doc commits to a wedge, gate confirms it

---name: researchdescription: Composite shortcut for research-dive — runs SENSE + DEFINE + DISCOVER, no BUILD/SHIP. For "understand before commit" mode.---You are the RESEARCH composite shortcut — research-dive mode pre-baked.## What this skill doesResearch-dive workflow: explore + understand, no code yet. Equivalent to:```/li:cycle --mode research-dive```Runs 3 phases (SENSE → DEFINE → DISCOVER). Skips PLAN/BUILD/REVIEW/SHIP/CAPTURE.Output: APPROVED design doc + discover-report.md. Operator can later resume with `/li:resume` to PLAN if research validates the direction.## When to use- New domain / unfamiliar territory- Pre-engagement "what do we have for X?" survey- Customer asks "what would this look like?" — design + discovery before commit- Operator wants to understand existing patterns before extending- Cost expectation: ~10-20k tokens, 20-40 min## When NOT to use- Ready to build — use `/li:cycle` (full)- Known territory — skip DISCOVER, just `/li:define`- Just want a quick lookup on a single topic — a focused web search is enough- No design discussion needed — use `/li:discover` standalone## Workflow### Step 1 — Pre-flightAskUserQuestion (brief): "Research mode produces design doc + discover-report but no code. Continue? (Y/n)"### Step 2 — Delegate to /li:cycle```bash/li:cycle --mode research-dive```Mode preset handles:- audience=solo- voice_tier=internal- compliance=none (read-only, no compliance gates apply)- Cost expectation pre-set medium### Step 3 — Post-research outputAfter DISCOVER DONE, surface:```RESEARCH COMPLETE — <wedge>Artifacts produced: - Design doc: <.claude/engineering/design-archive/lintel-*-design-*.md> (APPROVED) - Discover report: <.claude/runtime/state/discover-report-*.md>Findings: - ADRs surfaced: <N> - Lessons applied: <N> - Existing skills overlap: <N> - Recommended agents for future PLAN: <list>Next options: • /li:resume → continue to PLAN if direction validated • /li:capture → write findings as standalone artifact (no full cycle) • Decide later — design doc + discover-report persist```## Pause-points- Initial mode-confirmation- DEFINE forcing questions (all 6, per office-hours discipline)- Premise check + alternatives + design doc approval gate## IntegrationDelegates to `/li:cycle --mode research-dive`.## Anti-patterns- **Research-then-immediate-build without re-running PLAN** — research validates direction, PLAN turns it into tasks- **Skipping the design doc approval gate because "this is just research"** — design doc commits to a wedge, gate confirms it

Frequently asked questions

Research