Gsd Research Phase

B. AskUserQuestion → request_user_input Mapping

GSD workflows use AskUserQuestion (Claude Code syntax). Translate to Codex request_user_input:

Parameter mapping:

• header → header
• question → question
• Options formatted as "Label" — description → {label: "Label", description: "description"}
• Generate id from header: lowercase, replace spaces with underscores

Batched calls:

• AskUserQuestion([q1, q2]) → single request_user_input with multiple entries in questions[]

Multi-select workaround:

• Codex has no multiSelect. Use sequential single-selects, or present a numbered freeform list asking the user to enter comma-separated numbers.

Execute mode fallback:

• When request_user_input is rejected (Execute mode), present a plain-text numbered list and pick a reasonable default.

C. Task() → spawn_agent Mapping

GSD workflows use Task(...) (Claude Code syntax). Translate to Codex collaboration tools:

Direct mapping:

• Task(subagent_type="X", prompt="Y") → spawn_agent(agent_type="X", message="Y")
• Task(model="...") → omit (Codex uses per-role config, not inline model selection)
• fork_context: false by default — GSD agents load their own context via <files_to_read> blocks

Parallel fan-out:

• Spawn multiple agents → collect agent IDs → wait(ids) for all to complete

Result parsing:

• Look for structured markers in agent output: CHECKPOINT, PLAN COMPLETE, SUMMARY, etc.
• close_agent(id) after collecting results from each agent

Note: This is a standalone research command. For most workflows, use $gsd-plan-phase which integrates research automatically.

Use this command when:

• You want to research without planning yet
• You want to re-research after planning is complete
• You need to investigate before deciding if a phase is feasible

Orchestrator role: Parse phase, validate against roadmap, check existing research, gather context, spawn researcher agent, present results.

Why subagent: Research burns context fast (WebSearch, Context7 queries, source verification). Fresh 200k context for investigation. Main context stays lean for user interaction.

Normalize phase input in step 1 before any directory lookups.

0. Initialize Context

INIT=$(gsd-sdk query init.phase-op "{{GSD_ARGS}}")
if [[ "$INIT" == @file:* ]]; then INIT=$(cat "${INIT#@file:}"); fi

Extract from init JSON: phase_dir, phase_number, phase_name, phase_found, commit_docs, has_research, state_path, requirements_path, context_path, research_path.

Resolve researcher model:

RESEARCHER_MODEL=$(gsd-sdk query resolve-model gsd-phase-researcher --raw)

1. Validate Phase

PHASE_INFO=$(gsd-sdk query roadmap.get-phase "${phase_number}")

If found is false: Error and exit. If found is true: Extract phase_number, phase_name, goal from JSON.

2. Check Existing Research

ls .planning/phases/${PHASE}-*/RESEARCH.md 2>/dev/null

If exists: Offer: 1) Update research, 2) View existing, 3) Skip. Wait for response.

If doesn't exist: Continue.

3. Gather Phase Context

Use paths from INIT (do not inline file contents in orchestrator context):

• requirements_path
• context_path
• state_path

Present summary with phase description and what files the researcher will load.

4. Spawn gsd-phase-researcher Agent

Research modes: ecosystem (default), feasibility, implementation, comparison.

<research_type>
Phase Research — investigating HOW to implement a specific phase well.
</research_type>

<key_insight>
The question is NOT "which library should I use?"

The question is: "What do I not know that I don't know?"

For this phase, discover:
- What's the established architecture pattern?
- What libraries form the standard stack?
- What problems do people commonly hit?
- What's SOTA vs what the agent's training thinks is SOTA?
- What should NOT be hand-rolled?
</key_insight>

<objective>
Research implementation approach for Phase {phase_number}: {phase_name}
Mode: ecosystem
</objective>

<files_to_read>
- {requirements_path} (Requirements)
- {context_path} (Phase context from discuss-phase, if exists)
- {state_path} (Prior project decisions and blockers)
</files_to_read>

<additional_context>
**Phase description:** {phase_description}
</additional_context>

<downstream_consumer>
Your RESEARCH.md will be loaded by `$gsd-plan-phase` which uses specific sections:
- `## Standard Stack` → Plans use these libraries
- `## Architecture Patterns` → Task structure follows these
- `## Don't Hand-Roll` → Tasks NEVER build custom solutions for listed problems
- `## Common Pitfalls` → Verification steps check for these
- `## Code Examples` → Task actions reference these patterns

Be prescriptive, not exploratory. "Use X" not "Consider X or Y."
</downstream_consumer>

<quality_gate>
Before declaring complete, verify:
- [ ] All domains investigated (not just some)
- [ ] Negative claims verified with official docs
- [ ] Multiple sources for critical claims
- [ ] Confidence levels assigned honestly
- [ ] Section names match what plan-phase expects
</quality_gate>

<output>
Write to: .planning/phases/${PHASE}-{slug}/${PHASE}-RESEARCH.md
</output>

Task(
  prompt=filled_prompt,
  subagent_type="gsd-phase-researcher",
  model="{researcher_model}",
  description="Research Phase {phase}"
)

5. Handle Agent Return

## RESEARCH COMPLETE: Display summary, offer: Plan phase, Dig deeper, Review full, Done.

## CHECKPOINT REACHED: Present to user, get response, spawn continuation.

## RESEARCH INCONCLUSIVE: Show what was attempted, offer: Add context, Try different mode, Manual.

6. Spawn Continuation Agent

<objective>
Continue research for Phase {phase_number}: {phase_name}
</objective>

<prior_state>
<files_to_read>
- .planning/phases/${PHASE}-{slug}/${PHASE}-RESEARCH.md (Existing research)
</files_to_read>
</prior_state>

<checkpoint_response>
**Type:** {checkpoint_type}
**Response:** {user_response}
</checkpoint_response>

Task(
  prompt=continuation_prompt,
  subagent_type="gsd-phase-researcher",
  model="{researcher_model}",
  description="Continue research Phase {phase}"
)