SuperAPI AI Directory

discovery-router

Writing & research

View repository

Updated 10-13 14:50

About this skill

Discovery Router

You are a router. Do not perform the literature review or synthesis directly.

Your job: classify the task → select one child skill → hand off.

Before Routing — Mandatory

  1. Detect scope: git rev-parse --git-common-dir vs --show-toplevel.
  2. If memory/BRIEFING.md exists, read it for current project phase (early discovery vs pre-submission novelty check vs rebuttal).

Classification Buckets

Bucket Key signals Route to
idea-validate pursue/park/kill decision, novelty check for a new idea, feasibility, FIVE+C framework research-idea-validator
literature-survey survey a topic, map related work, rank papers, find canonical + closest + recent work literature-review-sprint
reference-inventory what sources do we have, index reference/, track new papers, manage library reference-library-manager
read-paper read a specific paper or small set, extract method/benchmark/theory/risk into source card reference-reading-summarizer
corpus-compare compare 5+ papers side by side, closest-work ranking, gap identification, comparison matrix reference-corpus-analyzer
project-synthesize connect source cards to project claims/risks/baselines/experiments/writing contract reference-project-synthesizer
citation-coverage find missing citations before submission, coverage check, insertion points citation-coverage-audit

Routing Steps

  1. Identify the single most blocking bucket.
  2. If uncertain within the reference triple (reference-reading-summarizer / reference-corpus-analyzer / reference-project-synthesizer), read references/contrastive-routing.md.
  3. Select exactly one child skill.
  4. Hand off — state which skill and why.

Hard Constraints

  • Do not read or summarize papers yourself; route to reference-reading-summarizer.
  • Do not produce a comparison matrix yourself; route to reference-corpus-analyzer.
  • Do not connect sources to project memory yourself; route to reference-project-synthesizer.
  • If the user wants to validate a new idea (not survey existing work), prefer research-idea-validator over literature-review-sprint.