Literature Triangulation

Produce a skeptical, source-grounded map of the topic. The output should read like something a careful researcher would say after spending a day with the papers — not like a Wikipedia summary.

The core problem this skill solves

Two failure modes dominate literature synthesis, and both are worse than saying "I'm not sure":

1. Hallucinated consensus. Claude fills gaps in coverage with plausible-sounding training-data priors, producing a confident synthesis that nobody with actual paper access would endorse. This is the most dangerous failure because it sounds authoritative.
2. Surface-level coverage. Reading abstracts and titles without following citation trails, missing the paper that actually introduced the method everyone else is citing, treating a blog post's characterization as authoritative, or stopping at the first three search results.

The antidote to both is triangulation: a claim earns confidence only when independent sources (ideally with different authors, institutions, and incentives) agree on it, and it holds on benchmarks the original authors didn't control.

Workflow

1. Clarify scope before searching