research-respond

name	research-respond
description	Checks and retrieves completed research results from Gemini Deep Research. Triggers: "fulfill a research request", "check research results", "poll Gemini for research results".
version	2
tier	lightweight

Checks and retrieves completed research results from Gemini Deep Research. [!!!] CRITICAL BOOT SEQUENCE:

1. LOAD STANDARDS: IF NOT LOADED, Read ~/.claude/directives/COMMANDS.md, ~/.claude/directives/INVARIANTS.md, and ~/.claude/directives/TAGS.md.
2. GUARD: "Quick task"? NO SHORTCUTS. See ¶INV_SKILL_PROTOCOL_MANDATORY.
3. EXECUTE: FOLLOW THE PROTOCOL BELOW EXACTLY.

⛔ GATE CHECK — Do NOT proceed to Phase 1 until ALL are filled in:

Output this block in chat with every blank filled:

Boot proof:

• COMMANDS.md — §CMD spotted: ________
• INVARIANTS.md — ¶INV spotted: ________
• TAGS.md — §FEED spotted: ________

[!!!] If ANY blank above is empty: STOP. Go back to step 1 and load the missing file. Do NOT read Phase 1 until every blank is filled.

Research Respond Protocol (Utility Command)

Lightweight — no log, no plan. Finds an open research request, calls Gemini, posts the result.

1. Setup

1. Template: The ~/.claude/skills/research-respond/assets/TEMPLATE_RESEARCH_RESPONSE.md template is your schema.
2. Session Dir: You MUST already be in an active session (via §CMD_MAINTAIN_SESSION_DIR). If not, ask the user which session to use.

§CMD_VERIFY_PHASE_EXIT — Phase 1

Output this block in chat with every blank filled:

Phase 1 proof:

• Template loaded: ________
• Active session dir: ________

---

2. Discovery

Find open or abandoned research requests.

1. Execute: §CMD_FIND_TAGGED_FILES for both #needs-research and #claimed-research across sessions/.
   • #needs-research — never started. Will need full query composition + API call.
   • #claimed-research — started but abandoned (previous session died). The request file contains an ## Active Research section with the Interaction ID. Can resume by polling the existing interaction.
2. If none found: Report "No open research requests." and stop.
3. If one found: Present it to the user for confirmation.
4. If multiple found: Present the list with status (new vs abandoned). User picks one.
5. Read: Read the selected request document fully.

§CMD_VERIFY_PHASE_EXIT — Phase 2

Output this block in chat with every blank filled:

Phase 2 proof:

• Tag search executed: ________
• Request selected: ________
• Request document read: ________

---

3. Compose Query

Build the research prompt from the structured request.

1. Extract from the request document:
   • The Query (Section 1)
   • The Context (Section 2)
   • The Constraints (Section 3)
   • The Expected Output (Section 4)
   • The Previous Research Interaction ID (Section 6 — may be "None")
2. Compose: Synthesize these fields into a coherent research prompt. Include all constraints and context to guide Gemini's research. The prompt should read as a complete research brief.

§CMD_VERIFY_PHASE_EXIT — Phase 3

Output this block in chat with every blank filled:

Phase 3 proof:

• Sections extracted: ________
• Research prompt composed: ________
• Previous Interaction ID: ________

---

4. Execute Research

Call the Gemini Deep Research API.

1. Determine output path: [session-dir]/RESEARCH_RESPONSE_[TOPIC].md.tmp (temporary — will be replaced by the final document).
2. Execute (background):
   • If initial request (Interaction ID = None):

     engine research <output-path> <<'EOF'
     [composed research prompt]
     EOF
     

   • If follow-up (Interaction ID present):

     engine research --continue <interaction-id> <output-path> <<'EOF'
     [composed follow-up prompt]
     EOF
     

   • Run with run_in_background: true.
3. Poll: Check the output file periodically (Read tool). When content appears, proceed.
4. Parse: Read the output file. Line 1 is INTERACTION_ID=<id>. Remaining lines are the report.

§CMD_VERIFY_PHASE_EXIT — Phase 4

Output this block in chat with every blank filled:

Phase 4 proof:

• research.sh executed: ________
• Output received: ________
• INTERACTION_ID parsed: ________
• Report content extracted: ________

---

5. Post Response

Write the response document and update the original request.

1. Create: Populate ~/.claude/skills/research-respond/assets/TEMPLATE_RESEARCH_RESPONSE.md:
   • Set Gemini Interaction ID from parsed output.
   • Set Original Request path.
   • Paste the full report into Section 2.
2. Write: Save as RESEARCH_RESPONSE_[TOPIC].md in the current session directory.
3. Tag: The template already includes #done-research in the Tags line.
4. Breadcrumb: Append to the original request file:

   engine log "$REQUEST_FILE" <<'EOF'
   ## Response
   *   **Responded By**: `[RESPONSE_FILE_PATH]`
   *   **Date**: [YYYY-MM-DD HH:MM:SS]
   *   **Gemini Interaction ID**: `[interaction-id]`
   *   **Status**: Complete
   EOF
   

5. Swap Tag: On the original request:

   engine tag swap "$REQUEST_FILE" '#needs-research' '#done-research'
   

6. Cleanup: Delete the .tmp output file if it exists.
7. Report: State "Research complete: [link]. Interaction ID: [id]." where [link] is a clickable link per ¶INV_TERMINAL_FILE_LINKS (Full variant).
8. Done: End turn. No debrief, no log. The response document IS the output.

§CMD_VERIFY_PHASE_EXIT — Phase 5 (PROOF OF WORK)

Output this block in chat with every blank filled:

Phase 5 proof:

• Response written: ________
• Breadcrumb appended: ________
• Tag swapped: ________
• .tmp cleaned: ________
• Completion reported: ________