deep-research-workflow

name	deep-research-workflow
description	Use when a research task benefits from an explicit phased workflow with structured source handoff across search, analysis, and writing, and when preserving citations across phases matters.
user-invocable	true
disable-model-invocation	false
metadata	{"openclaw":{"requires":{"bins":["bash"]}}}

Deep Research Workflow

This skill provides a phased research workflow:

1. search
2. analyze
3. write

Use it when the user wants a deeper research pass than a normal quick synthesis and when source preservation matters.

Required research loop

For every nontrivial research task, repeat this loop until the output is defensible:

1. Review — inspect the current scope, sources, claim mapping, and unresolved gaps.
2. Validate — check that claims still have source support, paper-like sources have a verification status, and observations are separated from inference.
3. Fix — repair missing evidence, narrow overbroad claims, add incomplete analysis when material scope remains unchecked, then repeat the loop.

Start with a short Research Brief for scope/evidence planning and end with a short Delivery Check before presenting the report as complete.

Runtime helper

Initialize a scaffold with:

exec: /workspace/skills/deep-research-workflow/run_deep_research_workflow.sh init --dir {{ PRIVATE_DATA_DIR }}/research

Verify the helper setup with:

exec: /workspace/skills/deep-research-workflow/run_deep_research_workflow.sh doctor

When to use

• deep topic research
• report-style synthesis
• research with explicit citation preservation
• tasks where search, interpretation, and final writing should be kept separate

Routing boundary

Prefer the normal browse-and-synthesize path for lightweight current-information lookups.

Prefer this skill when:

• the user wants an explicit phased workflow
• you need a structured handoff between search, analysis, and writing
• preserving source linkage across phases is part of the task quality bar

Workflow

Phase 1 — Search

• review the user request and write a compact Research Brief covering goal, scope, constraints, evidence plan, workflow, and risks
• gather relevant sources
• prefer primary sources when practical
• record source metadata with stable S1, S2, ... identifiers
• separate observed facts from tentative interpretations
• validate that each source record has enough metadata to be cited later
• fix the source ledger before moving to analysis

Use templates/deep-research-sources.md when helpful.

Zotero cross-check

Between Phase 1 and Phase 2, treat every paper-like source as a library-check task:

• search the local library with zotero
• assign exactly one verification status
• preserve that status in the source ledger

Allowed status values:

• [IN_LIBRARY]
• [NOT_IN_LIBRARY]
• [NOT_A_PAPER]
• [UNVERIFIED]

Phase 2 — Analyze

• review the source ledger and exclude only with an explicit reason
• group findings into themes
• identify conflicts, uncertainties, and gaps
• preserve source mapping for each important claim
• keep S* ids stable across all phases
• validate every important claim against one or more source ids
• fix unsupported claims by adding evidence, weakening the claim, or marking it unresolved

Detailed handoff structure:

• references/source-handoff.md
• templates/deep-research-analysis.md

Phase 3 — Write

• review the analysis matrix before drafting
• produce a structured output
• include only citations that survive from earlier phases
• distinguish observation, inference, and recommendation
• say incomplete analysis if material scope remains unchecked
• validate the draft with Review Findings before delivery
• fix any unsupported claim, scope drift, missing date, or overconfident wording
• finish with a visible Delivery Check that states readiness and remaining gaps

Output structure guidance:

• references/output-structure.md
• templates/deep-research-report.md

Skill handoffs

• Use docling before or between Phases 1 and 2 when local PDFs, HTML exports, or office documents need structure-aware parsing.
• Use paper-lookup during Phase 1 when external literature metadata or discovery is needed after the local library-first workflow.
• Use digest skills to seed Phase 1 when the task starts from tracked topics, alerts, or feeds.
• Use agent_group_discuss when a validation/fix cycle needs independent reviewers or adversarial critique.

Escalation rules

• Stay in this skill for single-agent phased deep research.
• Escalate to prose when the user explicitly wants structured multi-agent research-and-synthesis orchestration.
• Escalate to agent_group_discuss when the user wants panel-style discussion, debate, or multi-agent research perspectives.