problem-driven-literature-review

Problem-Driven Literature Review

Scope

Use this skill for literature review, related work, introduction background, research-gap synthesis, reference selection, and citation-role assignment in technical research papers, organized by a problem-driven protocol rather than a method-name list.

Read references/literature_review_protocol.md for the full protocol, worksheet, paragraph templates, reference-role table, reference metadata check, quality gates, and per-contribution-type examples before drafting or auditing a full review. For a short background paragraph, the principle and protocol summary below are enough.

Use another skill when the task is primarily:

• novelty risk, dangerous baselines, or whether the idea is defensible before writing: use idea-novelty-auditor (if available);
• manuscript storyline, section structure, contribution framing, or experiment-to-claim redesign: use paper-argument-reconstructor (if available);
• TeX-safe language polishing of the review text: use paper-polisher (if available).

Verification Mode

This skill has two modes. By default it writes or revises the review as described above. When the input is a related-work or background section another tool or skill has already drafted, switch to verification mode instead of regenerating:

• do not overwrite the other tool's wording;
• check it against this skill's rules: citation integrity, reference metadata quality, direct-competitor coverage, negative-claim hygiene, and whether the named structural gap is actually supported;
• report only the deviations and their locations; fix one in place only if it breaks a hard rule.

The suite's value is faithful, bounded output, not a competing rewrite. Verify rather than replace what a stronger generator already produced.

One-Sentence Principle

A literature review is not a list of what other people did. It should prove why the target paper is necessary:

S requires R.
Literature L solves parts of R under H.
H breaks under S, producing gap G.
We propose C to close G and validate it by V.