Forge Platform

M1 Forge Run Spine — Roadmap handoff

Translates docs/product-roadmap.md Release R1 into the first execution slice for sprint M1-forge-run-spine. Hermes, Factory, and OpenClaw are not in the critical path for M1 closure.

Product outcome

A static, reviewable chain proves the platform can represent intent lineage, approval, agent/workcell placeholders, LCDL trace placeholder, Fleet job placeholder, EvidencePacket, and decision-facing evidence — so Lenses (and other consumers) can attach without a parallel run model.

Why M1 now

R0 establishes credibility of boundaries and contract gaps; R1 is the lingua franca gate for every later release (Lenses visibility, LCDL traces, Fleet jobs, agents). Without R1, product teams will fork schemas.

Release alignment

  • Release: R1 — Forge Run Spine MVP
  • Milestones in scope for this handoff: R1.M1–R1.M6 (hardening + sample chain + Lenses handoff package)
  • Strategic objectives: S1 (credible spine), partial S2 (handoff to Lenses only)

Milestones included

ID Focus
R1.M1 ForgeRun v1 hardened
R1.M2 EvidencePacket v1 hardened
R1.M3 ApprovalRequest v1 hardened
R1.M4 AgentRun / WorkcellRequest / WorkcellResult hardened
R1.M5 Static sample chain validates
R1.M6 M2 Lenses handoff sample complete

Epics included

  • R1.E1EPIC-CORE-FORGE-RUN-SPINE
  • R1.E2EPIC-CORE-EVIDENCE-PACKET
  • R1.E3EPIC-LENSES-BLUEPRINTS-CONTEXT-PANEL (contract alignment / stub only in M1)
  • R1.E4EPIC-BLUEPRINTS-VERSONA-DEFINITIONS (references / stub)

Stories to draft next

Story Note
R1.S1 End-to-end static JSON bundle under samples/
R1.S2 EvidencePacket references all intersection-map artifact types (placeholders OK)
R1.S3 ApprovalRequest sample encodes human boundary per docs/approval-model.md
R1.S4 AgentRun + Workcell correlation IDs consistent with ForgeRun
R1.S5 M2 handoff doc: fixtures, glossary, panel mapping for Lenses

Sparks to draft next (immediate)

Spark Intent
R1.SP1 Audit schemas/*.json vs docs/intersection-map.md → gap list for R0.M3
R1.SP2 Normalize sample filenames / schema fields; keep scripts/check_foundation.py green
R1.SP3 Refresh docs/agent-workcell-model.md narrative for correlation
R1.SP4 Write evidence notes under sprints/M1-forge-run-spine/evidence/

Contracts to harden

  • schemas/forge_run.v1.schema.json
  • schemas/evidence_packet.v1.schema.json
  • schemas/approval_request.v1.schema.json
  • schemas/agent_run.v1.schema.json
  • schemas/workcell_request.v1.schema.json
  • schemas/workcell_result.v1.schema.json
  • (Placeholders in samples) schemas/lcdl_trace_summary.v1.schema.json, schemas/fleet_job_summary.v1.schema.json as referenced by EvidencePacket / intersection map

Do not change schemas on speculation — record inconsistencies as evidence requests for Architecture unless blocking.

Human gates

  • Schema breaking-change review if identifiers or required fields shift
  • Approval model alignment with docs/approval-model.md for ApprovalRequest samples
  • Platform assay per docs/milestone-plan.md before claiming M1 “close”

Evidence required to close M1

  • Green foundation check (python3 scripts/check_foundation.py)
  • Sample chain checklist or CI step (documented even if manual first)
  • sprints/M1-forge-run-spine/evidence/ updated with assay + sample pointers
  • R1.M6 handoff package consumable by Lenses team (path + README)

Risks

  • LCDL/Fleet naming drift in samples vs shipping repos — mitigate with explicit “placeholder” labels until R3/R4
  • Over-scoping M1 into Lenses UI — keep UI work in M2/R2

Suggested next prompts

  1. Architecture Versona: “Draft correlation-ID conventions for ForgeRun, ApprovalRequest, FleetJobSummary, WorkcellResult; flag schema impacts.”
  2. Forge Product Manager (authoring): “Decompose R1.S1–S5 into Ingots and Sparks with DoR per docs/sprint-structure.md.”
  3. Roadmap gate (PM): After M1 evidence is posted, run review template to produce roadmap-gate-review.md.