| name | TRIP-research |
| description | Exploratory research or spike - investigation without production code |
| disable-model-invocation | true |
| argument-hint | what do you want to investigate? |
Research Mode
You are now in research mode - for exploratory investigations that don't directly produce production code.
Use this for:
- Technology evaluation
- Feasibility studies
- Architecture exploration
- Performance investigation
- Bug root cause analysis
- Proof of concept
Your Task
Research: $ARGUMENTS
Step 0: Read fully @docs/ARCHI.md
Step 1: Define Scope
1.1 Clarify the Question
"What specific question(s) are we trying to answer?"
Document the research question(s) clearly:
- What do we need to find out?
- What would a successful outcome look like?
- What decisions will this research inform?
1.2 Compute Box
"How much thinking effort does this research require?"
Based on the complexity of the question(s), suggest an appropriate thinking level:
| Level | When to Use | Research Type |
|---|
| quick | Simple lookup, straightforward answer | "What's the syntax for X?" |
| brief | Minor analysis, single-source research | "How does library X handle Y?" |
| think | Standard research, comparing options | "Which library should we use for X?" |
| think hard | Complex analysis, architectural decisions | "How should we restructure module X?" |
| think harder | Deep investigation, multiple tradeoffs | "What's the best approach for X given constraints A, B, C?" |
| ultrathink | Critical decisions, extensive exploration | "Should we rewrite system X? What are all implications?" |
Use the AskUserQuestion tool to confirm the thinking level:
- Question: "Based on this research scope, I suggest using
[level] thinking. Does that seem appropriate?"
- Options: "Yes, use [level]" (proceed with suggested level), "Use a different level" (I want to adjust the thinking effort)
Once confirmed, the agent should apply the corresponding thinking effort throughout the investigation.
Step 2: Research Plan
Create a lightweight research plan (not a full TRIP plan):
# Research: [Topic]
## Question(s)
- [Primary question]
- [Secondary questions if any]
## Approach
1. [First thing to investigate]
2. [Second thing to investigate]
3. [...]
## Success Criteria
- [ ] [What we need to determine]
- [ ] [What we need to determine]
## Compute Level
[quick / brief / think / think hard / think harder / ultrathink]
Step 3: Confirm & Start
Present the research plan summary to the user, then use the AskUserQuestion tool with two questions:
Question 1 (header: "Start"):
- Question: "Research plan ready — Question: [primary question], Approach: [brief summary], Compute: [level]. Ready to start?"
- Options: "Yes, start research" (proceed with investigation), "Adjust the plan" (I have changes to the research scope or approach)
Question 2 (header: "Output"):
- Question: "How do you want the findings delivered?"
- Options: "Chat only" (respond directly in the conversation), "Write a memo" (create a file in
docs/6-memo/)
If "Adjust": Modify the plan based on user feedback, then re-present using AskUserQuestion.
If "Yes": Proceed with investigation. Remember the output preference for Step 5.
Step 4: Investigation
Conduct the research:
For Technology Evaluation
- Review documentation
- Check community/ecosystem health
- Look at alternatives
- Consider maintenance burden
- Assess learning curve
For Feasibility Study
- Identify constraints
- Prototype critical parts (throwaway code OK)
- Identify risks and unknowns
- Estimate effort
For Performance Investigation
- Establish baseline metrics
- Identify bottlenecks
- Test hypotheses
- Measure improvements
For Bug Investigation
- Reproduce the issue
- Trace the root cause
- Identify contributing factors
- Consider similar vulnerabilities
Step 4b: Codex Cross-Check
For decision-grade findings — architecture recommendations, technology choices, anything the user will build on (typically compute level think hard and above) — red-team the draft conclusion with the codex-ask skill before presenting. Skip for quick lookups.
export STATE_DIR=".claude/skills/codex-ask/state"
bash .claude/skills/codex-plan-review/scripts/start.sh \
--prompt-file .claude/skills/codex-ask/prompts/ask.tpl \
<topic-label> "Here is my draft recommendation: <summary + key rationale>. Red-team it: what am I missing, what would you choose instead, and why?"
Follow up in the same thread (resume.sh + followup.tpl) if the answer raises points worth probing. Then:
- Incorporate legitimate points into the findings (adjust the recommendation or add caveats).
- Record real disagreements in the memo's Open Questions section with both positions — the user decides.
- This is advisory, not gating: you own the final recommendation.
Step 5: Present Findings
If "Chat only" was selected
Present the findings directly in the conversation, structured as:
- Summary (2-3 sentences)
- Key Findings (numbered list)
- Recommendations (what to do, rationale, alternatives)
- Open Questions (if any)
Then use the AskUserQuestion tool:
- Question: "Research complete. What would you like to do next?"
- Options: "Elaborate on findings" (dive deeper into specific results), "Save as memo" (write findings to
docs/6-memo/ after all), "Plan implementation" (create a TRIP plan based on these results), "Done" (no further action needed)
If "Save as memo": write the memo using the template below, then confirm the file location.
If "Write a memo" was selected
Create findings document in docs/6-memo/:
File: docs/6-memo/research_[date]_[topic].md
# Research: [Topic]
**Date**: DD-MM-YYYY
**Author**: [Name]
## Summary
[2-3 sentence executive summary]
## Questions Investigated
### Q1: [Question]
**Finding**: [Answer]
**Confidence**: [High/Medium/Low]
**Evidence**: [What supports this conclusion]
### Q2: [Question]
[...]
## Key Findings
1. **[Finding 1]**: [Details]
2. **[Finding 2]**: [Details]
3. **[Finding 3]**: [Details]
## Recommendations
- **Recommended**: [What we should do]
- **Rationale**: [Why]
- **Alternatives Considered**: [What else was evaluated]
## Open Questions
- [Questions that remain unanswered]
- [Areas needing further investigation]
## Next Steps
- [ ] [Action item 1]
- [ ] [Action item 2]
## Appendix (optional)
### Code Snippets / Prototypes
[Any throwaway code created during research]
### References
- [Links to documentation, articles, etc.]
Then use the AskUserQuestion tool:
- Question: "Research documented at
docs/6-memo/research_[date]_[topic].md. What would you like to do next?"
- Options: "Elaborate on findings" (dive deeper into specific results), "Plan implementation" (create a TRIP plan based on these results), "Done" (no further action needed)
What This Workflow Produces
- Clear recommendations
- Basis for future TRIP planning
- Optionally: documented findings in
docs/6-memo/
What This Workflow Does NOT Produce
- No production code
- No version bump
- No changelog entry
- No commits to main
Transition to Implementation
If research concludes that implementation should proceed, use the AskUserQuestion tool:
- Question: "Based on this research, would you like me to create a TRIP plan for implementation?"
- Options: "Yes, create a plan" (use findings to inform
TRIP-1-plan), "Not yet" (I'll decide later)
If yes, use findings to inform TRIP-1-plan.