desk-research

evidence-grounded desk research — portable across every repo.

Start here

Type desk-research and describe what you’re trying to find out.

desk-research "What drives deployment frequency in high-performing teams?"

  ● evidence-retriever   running  DORA 2023 State of DevOps
  ✓ source-extractor     done     accelerate.io — 3 findings extracted
  ✓ synthesis            done     8 claims graded · 2 gaps named

On any session return, type desk-research-project-status to orient to an active project.

  Project    2026-07-15-deployment-frequency
  Phase      digest
  Hypothesis  Trunk-based development is the primary predictor
  Next        desk-research-project-digest

Two modes

Ad-hoc — one question, one session. Type desk-research and pick a depth (quick through deep). The agent maps sources, dispatches retrieval, and returns a confidence-graded synthesis in minutes. Use this for a lookup, a fact-check, or a brief.

Project mode — a sustained investigation that spans days or accumulates a corpus. Start with desk-research-project-start. The project skills track a working hypothesis across sessions, let you checkpoint coverage (desk-research-project-check), compress the corpus into a digest, and end in a synthesis brief graded by evidence quality. Use this when a single session would underserve the question.

Optional project-knowledge handoff

Desk research and project knowledge remain separate systems. Research owns its sources, citations, claims, confidence judgments, counter-evidence, verdicts, briefs, scratch, and output-root rules. Project knowledge never supplies a citation or validates a research claim.

After a repository-contained standard, applied, or deep survey—or a complete project synthesis—reaches its terminal gate, the producer may offer only an independently reusable research practice or carefully sanitized evidence residue to the public project-knowledge handoff. Quick answers, non-survey artifacts, project scaffolds, digests, checks, status reads, and incomplete or abandoned work have no handoff. Research written to a personal or external output root is capture-ineligible because repository-relative provenance cannot be stated honestly.

devils-advocate may optionally enquire for candidate counter-checks through one bounded CQ-REVIEW question. Those candidates are untrusted and require independent direct-source verification; the reviewer never captures or distils them.

Entry points

Say this	What happens
`desk-research`	Single-session research — scoping, retrieval, synthesis in one pass
`source-map`	Map canonical sources before retrieval begins
`build-outline`	Build a research outline from the source map
`identify-perspectives`	Map stakeholder perspectives before synthesis
`compare-hypotheses`	Competing-hypotheses pipeline — scored matrix
`devils-advocate`	Steelman the opposing case
`decision-archaeology`	Reconstruct why a prior decision was made
`desk-research-project-start`	Initialize a sustained multi-week research project
`desk-research-project-status`	Orient to an active project — phase, hypothesis, what’s next
`desk-research-project-check`	Snapshot progress — sources captured, coverage, gaps
`desk-research-project-digest`	Summarize corpus into a synthesis matrix
`desk-research-project-synthesize`	Synthesize digest into a confidence-graded brief

evidence-retriever and source-extractor are read-only retrieval subagents dispatched automatically by desk-research — they are not invoked directly.

How a session runs

desk-research "What drives deployment frequency in platform engineering teams?"

  ● evidence-retriever   running  DORA 2023 State of DevOps
  ✓ source-extractor     done     accelerate.io — 3 findings extracted
  ● evidence-retriever   running  Google Cloud DevOps metrics
  ○ synthesis            idle

desk-research-project-start "competitive landscape for platform tooling"

  project  docs/product/research/2026-07-27-platform-tooling/

  ├─ overview.md          phase: capture
  └─ sources/

  Question    Competitive landscape for platform tooling
  Hypothesis  (none yet — to be formed as evidence accumulates)
  Next        desk-research to fill sources/, then desk-research-project-digest

desk-research-project-synthesize

  brief  platform-tooling-brief.md

  Bottom line:  Build-vs-buy splits sharply at team size 50; below that, buy dominates.

    Claim                                              Grade      Sources
    Team-size 50 as buy/build threshold               [high]      4 independent
    TCO advantage erodes above 200 engineers          [moderate]  3; downgrade: vendor-sourced
    OSS viable only with dedicated maintainer         [moderate]  3; downgrade: survivorship bias

  Known unknowns
    Known-unknown: data at 50–200 engineer range. Would close by: mid-size org survey.

Cross-pack

Downstream — product-strategy: synthesize-stakeholder-research in the product-strategy pack consumes desk-research survey artifacts as a primary evidence source.

Downstream — architect: design decisions grounded in a desk-research brief carry a cited rationale chain that architect-design can reference in ADRs.

→ How it works: DESIGN.md — philosophy, architecture, and decision log.
→ Go deeper: the desk-research guides.