- Run
/startand/research-idea - Input: topic hints, constraints, and target venue direction
- Define topic, hypothesis, target audience, and success metrics
- Output:
research/ideas/andresearch/hypotheses/artifacts
- Run
/setup-researchclaw - Input: AutoResearchClaw install path and execution environment
- Ensure
researchclawis available with validated config - Standard minimal test command order:
- If
config.arc.yamlis missing:researchclaw init - Export or confirm the API key environment variable required by
llm.api_key_env - Run
researchclaw doctor - Run
researchclaw validate - Proceed to the next workflow step only after both checks pass
- If
- Output: reproducible runtime config snapshot
- Run
/citation-auditand/data-intake - Run
/data-cleanafter source freeze - Input: source list, acquisition constraints, and raw datasets
- Output: curated data, lineage manifest, and citation-quality report
- Run
/research-runand/analysis-run - Input: curated datasets, run config, and metric definitions
- Keep run logs, metrics definitions, uncertainty reports, and robustness notes
- Output:
artifacts/runs and analysis summaries underresearch/results/
- Run
/figure-build,/figure-qc, and drafting workflows - Run
/paper-style-sanity(warn-only) and/paper-gate-check - Input: analysis outputs, figure specs, and manuscript draft
- Output: traceable figures, coherent draft, gate report (PASS/CONCERNS/FAIL)
- Run
/repro-auditand/review-report - Run
/latex-venue-packand/latex-build-check - Input: gated draft, figure bundle, bibliography, and venue requirements
- Output: venue-ready submission package in
deliverables/submission/
- PASS: all required artifacts and checks are complete
- CONCERNS: submission possible with documented risks
- FAIL: critical integrity/reproducibility blockers remain unresolved
- Conversation templates:
docs/examples/ - Stable checklists and conventions:
docs/research-reference/