Skip to main content
  • Status: accepted · Date: 2026-07-17 · Deciders: Davor Runje

​
Context

Literature work spans generative idea-scouting (mining forward citations) and defensive positioning (related-works synthesis for a committed claim/paper). Two research passes examined whether these are one method or two, and whether they split by level.

​
Decision drivers

  • Both rest on one citation-graph substrate (OpenAlex + Semantic Scholar).
  • Within each function, hypothesis vs paper level is a parameter (ranking, depth, stopping), not a skill boundary.

​
Considered options

  1. One literature skill, two modes (scout, position), each level- parameterized.
  2. Two separate skills (lit-scout, lit-position).
  3. Per-level skills (hypothesis-literature, paper-literature).

​
Decision

Option 1. Shared graph plumbing; modes differ by intent/output; a level parameter tunes each.

​
Consequences

  • One methodology home; no duplicated graph plumbing.
  • Modes are distinct front-ends; the shared substrate is the backend.

​
Rejected alternatives

  • Two skills — duplicates the substrate for a cosmetic split.
  • Per-level skills — splits one body of methodology for form, not substance.
sub-spec 2 §1–§3; digests citation-scouting.md, related-works-synthesis.md.