/research-paper
/research-paper — Assembles a top-tier paper draft (abstract through conclusion plus references) from the synthesis, design, and analysis, with every citation verified to resolve.
Role: Principal Investigator (authoring) Pipeline position: mid-chain
Assembles a top-tier paper draft from the synthesis, design, and analysis — abstract, introduction, related work, method, results, discussion, limitations, conclusion, references. Every citation is verified to resolve to a real source; no phantom citations.
Canonical invocation
/research-paperWith arguments:
/research-paper [--suite-name NAME]Inputs
_inputs/synthesis.md (SOTA map + gap), _inputs/study-design.md (method), and _outputs/analysis.md (results).
Outputs
The paper deliverable at a host-natural location (paper/) — abstract, introduction, related work, method, results, discussion, limitations, conclusion, references.
The stage also writes the validation gate record. It holds the outcome for each reference and the count of resolved citations against the count claimed:
_outputs/paper-attestation.mdDownstream
/research-review (downstream consumer)
Workflow phases
The command follows the standard /research pipeline workflow:
- Load context — Read the synthesis, the study design, and the analysis.
- Verify the Sequence Gate — Confirm
_outputs/analysis.mdexists; otherwise halt withBlocked: run /research-analysis first. - Assemble the draft — Compose abstract through conclusion; the introduction and related work draw on the synthesis, the method on the design, the results on the analysis.
- Verify citations — Every reference resolves to a real source.
- Update the handoff manifest — Record the paper deliverable path.
- Emit the paper — the manuscript at the host-natural
paper/location.
Rigor floor
Under a double-blind venue the manuscript carries only what the anonymity policy permits: author, institution, funding, and the system under study are described in natural domain language without naming them, held out to separately-submitted metadata for a designated de-anonymization stage. The depth of the reproducibility and statistical reporting scales to the target-venue ambition.
Failure modes
| Symptom | Cause | Recovery |
|---|---|---|
| Blocked: run /research-analysis first | No analysis | Run /research-analysis to produce _outputs/analysis.md |
| Phantom citation | Reference does not resolve | Remove or correct the citation; never fabricate a source |
| Results overstated | Discussion exceeds the analysis | Constrain claims to the analyzed effect sizes + CIs |
Examples
# Assemble the paper draft from synthesis + design + analysis
/research-paper --suite-name <name>Cross-references
/research-analysis
/research-analysis — Analyzes the raw data per the preregistered plan: planned tests, effect sizes with confidence intervals, robustness checks, and disclosed deviations.
/research-review
/research-review — Peer-review-grade adversarial critique of the paper: a seven-axis reviewer scorecard (novelty, rigor, reproducibility, clarity, ethics, statistical review, reporting-guideline conformance) plus a severity-triaged required-revision list.