メニュー
Skill for running the hypothesis → experiment → finding workflow in scientific investigations. Covers pre-registration, sensitivity analysis, data integrity, review gates, replication, and serendipity handling.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | investigation-protocol |
| description | Skill for running the hypothesis → experiment → finding workflow in scientific investigations. Covers pre-registration, sensitivity analysis, data integrity, review gates, replication, and serendipity handling. |
| user-invocable | true |
| disable-model-invocation | false |
Hypothesis → Experiment → Finding workflow for scientific investigations.
Use this skill when working on DISCOVER or PROVE tasks that require evidence-based answers.
bd create "Test whether [hypothesis]" -t task -p 1 --parent <epic>
bd update <id> --description "TASK_TYPE:investigation | MODE:<discover|prove> | RIGOR:<adaptive|scientific|experimental>"
bd update <id> --notes "HYPOTHESIS: [specific, falsifiable statement]"
bd update <id> --notes "METHOD: [experimental approach]"
bd update <id> --notes "CONTROLS: [control conditions]"
bd update <id> --notes "EXPECTED_RESULT: [quantified expected outcome]"
bd update <id> --notes "CONFOUNDS: [identified confounding variables]"
bd update <id> --notes "STAT_TEST: [statistical test, alpha level]"
bd update <id> --notes "SAMPLE_SIZE: [N per condition with justification]"
Before running ANY experiment:
bd update <id> --notes "POWER_ANALYSIS: effect=[d] | alpha=[a] | power=[p] | min_N=[n] | method=[tool]"
bd update <id> --notes "SENSITIVITY_PLAN: param1=[range]; param2=[range]"
bd update <id> --acceptance "[specific criteria for confirm/refute]"
Investigation CANNOT proceed until:
# Methodologist approves the design
bd update <id> --notes "METHODOLOGIST_REVIEW: APPROVED | REVIEWER:<reviewer-id>"
# Execute the experiment
# Commit raw data to worktree: data/raw/
# Compute data hash
shasum -a 256 data/raw/<datafile>.csv
bd update <id> --notes "DATA_FILE:data/raw/<datafile>.csv"
bd update <id> --notes "DATA_HASH:sha256:<hash>"
Test at least 2 parameter variations:
bd update <id> --notes "SENSITIVITY: param1=[values] -> effect=[range] | ROBUST:yes|no"
bd update <id> --notes "SENSITIVITY: param2=[values] -> effect=[range] | ROBUST:yes|no"
bd create "FINDING: [result summary with CI]" -t task --parent <epic>
bd update <finding-id> --notes "TYPE:finding | VALENCE:positive|negative|inconclusive | CONFIDENCE:0.X"
bd update <finding-id> --notes "SOURCE_TASK:<investigation-task-id>"
bd update <finding-id> --notes "EFFECT_SIZE:[d] | CI_95:[lo, hi] | N:[n] | STAT_TEST:[test] | P:[p]"
bd update <finding-id> --notes "DATA_FILE:<path>"
bd update <finding-id> --notes "DATA_HASH:sha256:<hash>"
bd update <finding-id> --notes "SENSITIVITY: [summary] | ROBUST:yes|no"
bd update <finding-id> --notes "REPLICATED:no | STAT_QUALITY:[pending] | REVIEWED_BY:[pending]"
Findings must pass through these gates before entering the knowledge store:
| Gate | Rigor Level | Reviewer |
|---|---|---|
| Statistical review | Analytical+ | Statistician |
| Adversarial review (partially blinded) | Scientific+ | Critic |
| Consistency check | Scientific+ | Synthesizer |
Negative findings are equally important:
bd create "FINDING: [variable] has no measurable effect on [outcome]" -t task --parent <epic>
bd update <id> --notes "TYPE:finding | VALENCE:negative | CONFIDENCE:0.X"
bd update <id> --notes "IMPLICATION: [what this rules out, how it narrows hypothesis space]"
bd create "Replicate: [original finding description]" -t task -p 2 --parent <epic>
bd update <id> --description "TASK_TYPE:replication | RIGOR:scientific | TARGET:<original-task-id>"
bd update <id> --notes "MUST use different implementation than original"
bd update <id> --notes "MUST use same hypothesis and acceptance criteria as original"
When you discover something unexpected:
bd update <id> --notes "SERENDIPITY:HIGH | DESCRIPTION:[what you found]"
The orchestrator may allocate up to 2 OODA cycles to pursue the lead.
<worktree>/
├── src/ # Code (if build task)
├── experiments/ # Experiment scripts
├── data/
│ ├── raw/ # Raw experimental data (CSV, JSON)
│ ├── processed/ # Cleaned/transformed data
│ └── figures/ # Generated plots
└── report.md # Structured summary of work done
Skill for using Beads (bd) as the single source of truth for task tracking, dependency management, and agent coordination. Covers all bd commands needed for multi-agent orchestration.
Skill for working with artifact contracts — the file-level dependency system that prevents pipeline failures by ensuring tasks produce expected outputs and only start when required inputs exist.
Skill for generating scientific figures from experimental data. Ensures every figure referenced in LaTeX documents exists on disk before compilation, preventing blank PDFs and broken references.
Complete recipe for dispatching, monitoring, merging, and cleaning up worker agents. Read this skill BEFORE dispatching any worker. Covers build_worker_prompt, spawn-agent.sh, merge-agent.sh, checkpoint updates, and common failure modes.
Use when asked to reduce code complexity, simplify code, refactor safely, merge micro-files, remove dead code, shrink functions, reduce coupling, clean modules, or improve maintainability in src/voronoi/.
Use for adversarial review of Voronoi plans, findings, claims, manuscripts, and final deliverables. Covers reviewer-style objections, confounds, alternative explanations, evidence gaps, and falsification-oriented critique.