Literature review

Evidence-Grounded Literature Review

Purpose

Audit and synthesize a retrieved literature corpus against a research project.

When To Use

Use after paper-discovery has produced a retrieval inventory or when the user supplies a current corpus with provenance. This skill owns literature audit, source-grounded synthesis, gap analysis, and prior-art comparison.

Use report-arbitration for conflicting reports or syntheses. Use research-manuscript-audit for draft audit. Use paper-discovery for missing, stale, or incomplete retrieval.

Required Inputs

• {{PROJECT_NAME}}
• {{RESEARCH_QUESTION}}
• {{DOMAIN}}
• {{TARGET_AUDIENCE_OR_VENUE}}
• {{PROJECT_ARTIFACTS}}
• {{RETRIEVED_CORPUS}}: paper-discovery inventory or equivalent source set with provenance.
• {{SEARCH_AXES}}
• {{DOMAIN_SPECIFIC_EVIDENCE_SCHEMA}}
• {{AS_OF_DATE}}

Optional Inputs

• {{SOURCE_TEXTS_OR_PDFS}}
• {{CONSTRAINTS}}
• {{MEASUREMENT_VALIDATION_REQUIREMENTS}}
• {{SAFETY_PRIVACY_OR_RELEASE_CONTEXT}}
• {{SHARED_RESEARCH_CONTRACT}}
• {{AVAILABLE_TOOLS}}
• {{OUTPUT_LOCATION}}

Authority Order

Use {{SHARED_RESEARCH_CONTRACT}} when supplied. Otherwise apply this order by claim type:

1. Explicit user instructions, release constraints, safety constraints, and local project conventions.
2. Raw or near-raw project evidence for project claims.
3. Primary external sources for external claims.
4. Reproducible analyses and transparent secondary sources that cite primary artifacts precisely.
5. Canonical current-state project documents and authority ledgers.
6. Dated archival records and historical notes.
7. Input reports, model outputs, comments, and prior narratives, treated as leads.
8. Analyst judgment, labeled as judgment.

Tool And File-Access Assumptions

Use local files, supplied inventories, citation metadata already present in the inventory, and browsing when available for verification. Scholarly retrieval belongs to paper-discovery; route missing or stale corpus coverage there.

If browsing, files, or artifact creation are unavailable, state the limitation and produce a concrete verification backlog. If {{OUTPUT_LOCATION}} is supplied and artifact creation is available, write the review there.

Handoffs

Upstream artifacts are paper-discovery Result Inventory, Search Protocol, Citation Expansion, Licensing And Reuse Notes, Retraction And Hygiene Flags, and Verification Backlog.

Downstream artifacts are Literature Map, Source Evidence Matrix, Claim Ledger, Closest Prior Art, Gap Statement, Contribution Options, Evidence Roadmap, and Open Checks.

Pass conflicts across reports to report-arbitration. Pass accepted synthesis and claim constraints to clean-room-framing-options or research-manuscript-audit. Pass durable findings, stale claims, and evidence links to update-docs.

Claim-Support Statuses

Use one status per nontrivial claim:

• Supported
• PartiallySupported
• ReasonableInference
• Ambiguous
• Overstated
• Unsupported
• StaleOrSuperseded
• Unverified
• Opinion

Process

1. Inspect {{PROJECT_ARTIFACTS}} before making project claims.
2. Check {{RETRIEVED_CORPUS}} for search protocol, source provenance, inclusion criteria, exclusion criteria, date coverage, licensing, and retraction hygiene.
3. Put absent or stale retrieval coverage in Verification Backlog and route retrieval work to paper-discovery.
4. Screen sources into Must Read, Useful Context, Peripheral, and Exclude.
5. Extract structured evidence using {{DOMAIN_SPECIFIC_EVIDENCE_SCHEMA}}.
6. Add measurement-validation fields when {{MEASUREMENT_VALIDATION_REQUIREMENTS}} applies.
7. Separate source claims, validated findings, derived metrics, broader interpretations, uncertainty, and judgment.
8. Assign one claim-support status to every nontrivial claim.
9. Cluster sources into methodological or conceptual families.
10. Compare the project against closest prior work with exact overlap and difference dimensions.
11. Identify literature gaps, tooling gaps, measurement gaps, data gaps, validation gaps, and theory gaps.
12. Produce conservative, strong, and ambitious contribution options.
13. Recommend one primary path with evidence needed, risks, and pivot criteria.

Output Contract

Produce a Markdown review with these sections:

1. Executive Answer: direct answer to the research question, confidence, and key caveats.
2. Project Snapshot: current project state grounded in artifacts.
3. Corpus Protocol Audit covering search coverage, screening criteria, exclusions, and verification backlog.
4. Literature Map: source families and how they relate to the project.
5. Source Evidence Matrix covering source, artifact type, method, finding, limits, relevance, and provenance.
6. Claim Ledger covering claim, support status, evidence, counterevidence, confidence, and corrected wording.
7. Grounded Findings: claims the corpus supports.
8. Disagreements And Uncertainties covering conflicts, stale areas, fragile assumptions, and open checks.
9. Closest Prior Art: comparison matrix with overlap and difference dimensions.
10. Gap Statement: narrow defensible gap and required caveats.
11. Contribution Options: conservative, strong, and ambitious paths.
12. Recommended Path: evidence needed, risks, and pivot criteria.
13. Evidence Roadmap covering observations, analyses, sources, or validations that would change the plan.
14. Risk, Privacy, Safety, Or Release Checks: include when relevant.
15. Downstream Handoff with artifacts for report-arbitration, clean-room-framing-options, research-manuscript-audit, or update-docs.

Quality Gate

Before finalizing, verify:

• The corpus came from paper-discovery or another supplied source set with provenance.
• Missing retrieval appears in Verification Backlog.
• Every nontrivial claim has evidence, a support status, and an uncertainty label where needed.
• Novelty claims identify closest prior art and exact comparison axis.
• Metrics, instruments, proxies, and benchmarks support claims only at their validated scope.
• Citation use matches the exact claim made.
• Tool, browsing, file, and artifact claims match completed actions.