Epic vs Product Spark vs Forge Spark (dual profile)

The collision

Agents mint Forge Sparks on Charge in Epic-profile repos, treat Epics as Product Sparks, or build parallel WBS Task backlogs — broken OpenSpec traceability and duplicate Charge hierarchies.

How to choose

1) Detect profile — forge.config.yaml, docs/PROJECT.md, or Charge table (Active Sparks vs Active Epics). 2) Shippable increment / Assay / milestone M1? → Product Spark (both profiles). 3) Epic execution profile + ready Epic (OpenSpec size gate, one repo, reviewable diff)? → Charge Epic (M1E3) — not Forge Spark. 4) Epic execution profile + work inside Charged Epic? → L1–L2 runs — not Charge rows, not …T{n} WBS Tasks, not tasks.md merge gates. 5) Time-boxed learning only? → discipline spike (spike_discipline) — not Product Spark, Forge Spark, or Charged Epic. 6) Core Forge — today's executable task in Charge? → Forge Spark (phase-prefixed task id). 7) Core Forge — colloquial sizing only? → re-check steps 2 and 6; use product-spark-vs-forge-spark for Spark collisions.

Using several at once

Dual profile: core keeps Ore → Ingot → Forge Spark → Charge. Overlay switches Charge grain to Epics with OpenSpec 1:1. Do not imply Spark is deleted from Forge — teams choose profile per repo. WBS Epic IDs align with committed Epics under profile only.

Terms covered

• Epic-execution-profile
• Product-Spark
• Forge-Spark
• Charge
• WBS

Examples from chat / plan.md

forge-lenses under profile → Charge M1E3 with openspec/ change; agents classify → plan → apply inside Epic — no discover:foo Spark on Charge.

Core product repo → Charge lists discover:api-schema Forge Spark; WBS M1E1S1T1 aligns Task and Spark on one ID spine.

Bridge page epic-vs-spark-vs-product-spark — read when multiple abbreviations appear in one sentence.