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

Context

How to structure the lifecycle skills so hypothesis-level and paper-level are uniform, and so the “missing” stages (related-works, publish-decision) fit without proliferating skills.

Decision drivers

  • Object×action symmetry across levels.
  • A firewall: explore proposes, resolve disposes, synthesize reports.
  • The “missing” stages are analogs of existing staged docs, not new skills.

Considered options

  1. Two skills per level: generate (exploration) + resolve (testing/synthesis), with staged docs; “related works” = positioning (dual of strategy), “publish decision” = decision (dual of findings).
  2. One big lifecycle skill per level with sub-commands.
  3. A skill per lifecycle stage (many small skills).

Decision

Option 1. The two missing stages resolve into staged documents in the resolve skill, preserving the mirror.

Consequences

  • Minimal skill count; strong symmetry.
  • Related-works and publish-decision are documents/gates, not skills.

Rejected alternatives

  • One big skill — breaks the explore/dispose/report firewall; hard to reason about.
  • Skill-per-stage — over-fragmented; duplicated context.
meta-spec §3.1–§3.2; sub-spec 1 §1–§2.