Thomashighbaugh/sciomc

Research Skill

Orchestrate parallel scientist agents for comprehensive research workflows with optional AUTO mode for fully autonomous execution.

Usage

/sciomc <goal>                    # Standard research with user checkpoints
/sciomc AUTO: <goal>              # Fully autonomous until complete
/sciomc status                    # Check current research session status
/sciomc resume                    # Resume interrupted research session
/sciomc list                      # List all research sessions
/sciomc report <session-id>       # Generate report for session

Research Protocol

Stage Decomposition

Decompose research goals into 3-7 independent stages. Each stage has: Focus, Hypothesis (optional), Scope, Tier (LOW|MEDIUM|HIGH).

Parallel Scientist Invocation

Fire independent stages in parallel via Task tool. Always pass model parameter explicitly:

| Task Complexity | Agent | Model | Use For | |-----------------|-------|-------|---------| | Data gathering | scientist (model=haiku) | haiku | File enumeration, pattern counting, simple lookups | | Standard analysis | scientist | sonnet | Code analysis, pattern detection, documentation review | | Complex reasoning | scientist | opus | Architecture analysis, cross-cutting concerns, hypothesis validation |

@scientist(model="haiku", prompt="[RESEARCH_STAGE:1] Investigate...")
@scientist(model="sonnet", prompt="[RESEARCH_STAGE:2] Analyze...")
@scientist(model="opus", prompt="[RESEARCH_STAGE:3] Deep analysis of...")

Verification Loop

After parallel execution, cross-validate findings:

@scientist(model="sonnet", prompt="[RESEARCH_VERIFICATION] Cross-validate these findings for consistency. Check: 1) Contradictions between stages 2) Missing connections 3) Gaps in coverage 4) Evidence quality. Output: [VERIFIED] or [CONFLICTS:<list>]")

AUTO Mode

Runs complete research workflow autonomously with loop control. Max 10 iterations. Continues until [PROMISE:RESEARCH_COMPLETE] emitted or max iterations reached. Cancel via /cancel.

Promise Tags

• [PROMISE:RESEARCH_COMPLETE] — All stages done, verified, report generated
• [PROMISE:RESEARCH_BLOCKED] — Cannot proceed (missing data, access issues)

Loop Control

[RESEARCH + AUTO - ITERATION {{ITERATION}}/{{MAX}}]
Current state: {{STATE}} | Completed: {{COMPLETED_STAGES}} | Pending: {{PENDING_STAGES}}

Parallel Execution Patterns

Independent stages — fire simultaneously:

@scientist(model="haiku", prompt="[STAGE:1] Analyze src/api/...")
@scientist(model="haiku", prompt="[STAGE:2] Analyze src/utils/...")

Hypothesis battery — test multiple hypotheses in parallel:

@scientist(model="sonnet", prompt="[HYPOTHESIS:A] Test if caching improves...")
@scientist(model="sonnet", prompt="[HYPOTHESIS:B] Test if batching reduces...")

Cross-validation — sequential after all parallel stages complete:

@scientist(model="opus", prompt="[CROSS_VALIDATION] Validate consistency across all findings...")

Concurrency limit: Max 20 concurrent scientists. Batch if more.

Session Management

.opencode/state/research/{session-id}/
  state.json              # Session state and progress
  stages/                 # Stage findings
  findings/raw/           # Raw findings from scientists
  findings/verified/      # Post-verification findings
  report.md               # Final synthesized report

State file: { id, goal, status, mode, iteration, maxIterations, stages: [{ id, name, tier, status, findingsFile }], verification: { status, conflicts }, createdAt, updatedAt }

Tag Extraction

Scientists use structured tags. Extract with regex:

[FINDING:<id>] <title> ... [/FINDING]
[EVIDENCE:<finding-id>] File: <path>, Lines: <range> [/EVIDENCE]
[CONFIDENCE:HIGH|MEDIUM|LOW] <reasoning>
[STAGE_COMPLETE:<id>]
[VERIFIED|CONFLICTS:<list>]

Quality threshold: Each [FINDING] needs ≥1 [EVIDENCE], a [CONFIDENCE], absolute file paths, and must be reproducible.

Report Generation

Template: Executive Summary → Methodology (stages table) → Key Findings (with evidence) → Cross-Validation Results → Limitations → Recommendations → Appendix (raw data links, state.json).

Figures embedded via [FIGURE:path] Caption [/FIGURE] markers. Types: architecture diagrams, flow charts, dependency graphs, timelines, comparison tables.

Configuration

{ "hubs": { "research": { "maxIterations": 10, "maxConcurrentScientists": 5, "defaultTier": "MEDIUM", "autoVerify": true, "generateFigures": true, "evidenceContextLines": 5 } } }

Troubleshooting

• Stuck in verification loop? Check for conflicting findings between stages; re-run specific stages with different approach.
• Low-quality findings? Check tier assignment; ensure prompts include clear scope and expected output format.
• AUTO mode exhausted iterations? Review state to see where stuck; consider breaking into smaller sessions.
• Missing figures? Verify figures/ directory exists and [FIGURE:] paths are relative to session directory.