/research
Reference for the /research command: invocation, arguments, inputs and outputs.
The research pipeline wrapped as a single dynamic multi-agent workflow. One call drives a research mission from a raw idea to a disseminated, peer-reviewed paper by dispatching the first-class stages — research-ideate → research-spec → research-theory → research-sources → research-synthesis → research-proposal → research-design → research-experiment → research-analysis → research-paper → research-review → research-publish → research-disseminate — as workflow phases under named Handoff-Manifest return contracts, verifying each stage's hand-off refute-by-default before the next stage consumes it. Invoke it with a research question or domain ('research whether X causes Y', 'investigate the SOTA on Z', 'run the study from idea to paper'), an existing research suite path to resume, or --quick for a one-shot cited report. Stage logic stays in commands/research-*.md; this command adds only the workflow harness. Multi-agent dispatch and continuous chaining are opt-in and confirmation-gated.
Invocation
/research [<<question>> | suite path] [--autonomous] [--max-rounds N] [--quick] [--verify-panel N] [--from ideate|spec|theory|sources|synthesis|proposal|design|experiment|analysis|paper|review|publish|disseminate]The definition sets disable-model-invocation: true, so a harness that honors that key starts this command only when you type it.
Pipeline position
Wrapped-workflow meta-orchestrator over the whole /research pipeline; the canonical single-call entry for the pipeline. It consumes a research mission (raw question or an existing suite path) and drives the stage chain to completion through the workflow harness; it emits no artifact of its own beyond the chain's outputs (the suite, its per-stage artifacts, the paper deliverable, the publication record) plus the workflow's deterministic result surface and run trace.
Inputs
The operator's question or suite path; the --autonomous opt-in; the --quick mode; the --verify-panel N budget; the optional --from <stage> resume point. At each stage transition, the upstream stage's Handoff Manifest at {suite}/_inputs/handoff-manifest.yml (per src/apothem/schemas/handoff-manifest.yaml) is the named return contract the next phase consumes.
| Argument | Type | Required | Description |
|---|---|---|---|
<<question>> | suite path | String | Yes | A raw research question or domain (→ chain starts at research-ideate) OR an existing <project-root>/.apothem/plans/{suite}/ path (→ resume from the first incomplete stage). |
--autonomous | Flag | No | Opt into continuous chaining (no per-stage-boundary halt). Default: halt at each boundary for confirmation, per rules/agnostic-posture.md + rules/context-management.md §4A. Irreversible / outward-facing steps (data collection, publication submission, dissemination) stay per-action gated even under this flag. |
--max-rounds N | Integer | No | The per-stage remediation cap (default: 3). A hand-off still refuted after N remediation rounds halts the run as BLOCKED with the residual findings and their owning stages, per rules/planning-techniques.md §1. |
--quick | Flag | No | Run research-sources + research-synthesis only → a cited report with per-claim confidence bands. The single-shot research-engagement slice; bypasses the ideate → spec → theory upstream and the proposal → design → experiment → analysis → paper → review → publish → disseminate downstream stages. |
--verify-panel N | Integer | No | Refute-by-default critics per stage hand-off (default: 3). |
--from ideate|spec|theory|sources|synthesis|proposal|design|experiment|analysis|paper|review|publish|disseminate | Enum | No | Explicit resume stage (overrides the manifest-derived resume point). |
Outputs
The chain's own artifacts (owned by each stage command — the research spec, source ledger, synthesis, study design, experiment log, analysis, paper, review report, publication record), plus the workflow's deterministic result surface: the research outcome, the per-stage verified hand-offs with evidence, the per-deliverable advance-gate decisions with the thresholds each cleared (verified from the stage's own advancement-gate attestation, never recomputed), the disclosure ledger of any beyond-mission remediation, the fifteen-bar gate attestation, the per-run workflow trace (stages run, manifest hand-offs, verification verdicts, halt/continue decisions), and the single recommended next move.
Next step
Invoke /research <question> for a fresh research mission (the wrapped workflow starts at /research-ideate and drives to a disseminated, peer-reviewed paper), /research <question> --quick for a one-shot cited report (runs /research-sources + /research-synthesis only), or /research <project-root>/.apothem/plans/{suite}/ to resume an existing suite from its first incomplete stage; review each stage hand-off's refute-by-default verification, then pass --autonomous to chain continuously once the decomposition and return contracts read correctly. The planned (halt-at-boundary) mode is the safe default — autonomy is the explicit opt-in.
Source
Generated from src/apothem/commands/research.md, the command definition every harness installs.
/plan-status
/plan-status — Read-only plan-suite progress reporter — emits status without modifying any files.
/research-ideate
/research-ideate — Formulates the problem space from a domain seed: opportunity-and-gap scanning, invalidated-prior-hypothesis scanning, and ranked candidate research-question generation — the head stage of the /research pipeline, feeding /research-spec.