hyperresearch-5-depth-investigation

name	hyperresearch-5-depth-investigation
description	Step 5 of the hyperresearch V8 pipeline. Spawns K depth-investigator subagents in parallel (one per scored locus), each producing one interim note with a Committed Position section. Investigators read full source bodies for their locus and may fetch additional sources within their source_budget. Invoked via Skill tool from the entry skill (full tier only).

Step 5 — Depth investigation (parallel, K = len(loci))

Tier gate: SKIP entirely for light tier. Only full tier runs depth investigation.

Goal: produce ONE interim-{locus}.md note per locus with dense synthesis that the draft sub-orchestrators (step 10) will draft from.

---

Recover state

Read these inputs:

• research/scaffold.md — vault_tag
• research/loci.json — scored loci with source_budget per locus
• research/temp/contradiction-graph.json (if step 3 ran)
• research/query-<vault_tag>.md — canonical research query

---

Procedure

1. Spawn K hyperresearch-depth-investigator subagents in parallel (ONE message, all Task calls). One per locus with source_budget > 0, capped at 6.

   Spawn template:

   subagent_type: hyperresearch-depth-investigator
   prompt: |
     RESEARCH QUERY (verbatim, gospel):
     > {{paste research/query-<vault_tag>.md body}}
   
     QUERY FILE: research/query-<vault_tag>.md
   
     PIPELINE POSITION: You are step 5 (depth-investigator) of the
     hyperresearch V8 pipeline. Step 4's loci analysts produced research/loci.json;
     after you return, step 6 will reconcile your committed position against
     the other investigators' positions in research/comparisons.md.
   
     YOUR LOCUS (from research/loci.json):
     - name: "<locus name>"
     - one_line: "<one-line locus description>"
     - flavor: "dialectical" / "synthesis" / "technical"
     - source_budget: <integer from loci.json>
     - rationale: "<why this locus matters>"
   
     YOUR INPUTS:
     - corpus_tag: <vault_tag>
     - locus_name: <locus name>
     - source_budget: <hard cap on additional sources you can fetch>
   
     CRITICAL: Read the full source text of relevant vault notes (via
     `hyperresearch note show <id1> <id2> ... -j`) BEFORE writing your
     interim note. Drafting from summaries alone produces paraphrase;
     drafting from full text produces synthesis. Use your source_budget
     to fetch additional sources beyond the width corpus if needed.
   
     OUTPUT: Write a single interim note via the hyperresearch CLI with
     type=interim, tags = <vault_tag> + locus-<locus-name>. The note MUST
     end with a "## Committed position" section that takes a SIDE on the
     dialectical question (or a synthesis verdict for non-dialectical
     loci). Include calibration: confidence level, what evidence would
     change your mind.
   

   Each investigator's hard cap is locus.source_budget, not a flat number.

2. Each investigator writes ONE interim note into the vault with type: interim and tags <vault_tag> + locus-<locus-name>. Return value is the note id.

3. Wait for all K to complete. Investigators can fail independently. Proceed with whichever succeeded. If >50% failed, stop and reassess loci quality with the user.

4. Read the interim notes. After all return, list them:

   $HPR search "" --tag <vault_tag> --type interim --json
   

   Then batch-read them:

   $HPR note show <id1> <id2> ... -j
   

   Hold the Committed Position sections in your context — they are the load-bearing input to step 6 (cross-locus reconciliation).

INVARIANT: Every interim note ends with a ## Committed position section. An interim note ending with descriptive summary only is defective — flag it and re-spawn that investigator with the committed-position requirement emphasized.

---

Exit criterion

• One interim note per locus with source_budget > 0, each tagged <vault_tag> + locus-<locus-name>
• Every interim note ends with ## Committed position

If >50% of investigators failed: stop and escalate.

---

Next step

Return to the entry skill (hyperresearch). Invoke step 6:

Skill(skill: "hyperresearch-6-cross-locus-reconcile")