[Kernel v0] First useful vertical slice #3

New issue

Closed

opened 2026-09-22 16:46:21 +00:00 by kilo · 0 comments

kilo commented

2026-09-22 16:46:21 +00:00

Member

Goal

Coordinate the smallest deterministic Creatidy Kernel sequence that delivers a local durable Program with two dependent owner-approved engineering WorkUnits, deterministic recovery and acceptance semantics, replaceable resource/runtime/forge boundaries, and a reference flow that opens a Forgejo PR without automatic merge or deployment.

Why

A0 accepted the architecture and proved only the resource-boundary foundation. This record proves Kernel semantics before live integrations and keeps Forgejo, coding runtimes, and Scarcity Router outside core authority.

Scope

In scope

Coordinate implementation issues #4-#11 and #13.
Complete the first useful core path with FixedAllocator, local SQLite/artifacts, Forgejo, and one existing coding harness.
Treat #11 as an optional external-consumer ResourceAllocator proof after #10. The owner explicitly selected #11 for execution on 2026-09-27; the integration remains optional for Kernel users and does not transfer Program authority to Scarcity Router.

Final dependency graph:

#4 K1 Domain and immutable intent
  -> #5 K2A Durable SQLite history and projections
    -> #13 K2B External operation journal, outbox, and artifact durability
      -> #6 K3 Synthetic execution, workspace, and authority enforcement
        -> #7 K4 Evidence-based verification, acceptance, remediation, and outcomes
          -> #9 K5 Forge port and Forgejo adapter
          -> #8 K6 Runtime port and Codex app-server adapter
            #9 + #8 -> #10 K7 CLI reference slice and restart/recovery proof
              -> #11 K8 optional external Scarcity Router ResourceAllocator proof

Out of scope

REST, MCP, A2A, web UI, IDE integration, distributed workers/controllers, Temporal, Kubernetes, automatic merge, deployment, custom sandbox or secrets manager, custom policy DSL, adaptive allocation, Quota Governor, full Context Compiler, central telemetry, M5-B, Prefect, private migration, pilots, and releases.

This is a coordination/roadmap record only. It does not authorize Program Execution Mode, autonomous task selection, or merge, main, release, deployment, or issue-creation authority. Each implementation issue must still be explicitly selected by the owner.

Acceptance Criteria

Issues #4-#11 and #13 have executable scope, exclusions, acceptance evidence, exact numbered links, and architecture references.
#4-#7 and #13 prove deterministic and synthetic Kernel semantics without live Forgejo, Codex, Scarcity Router, M5-B, Prefect, or cloud services.
#10 drives a finite two-WorkUnit DAG through FixedAllocator, runtime, candidate retrieval, independent verification, acceptance, explicit successor input, restart/recovery, and Forgejo PR creation without automatic merge/deployment.
#11 repeats the accepted scenario as an optional external Scarcity Router ResourceAllocator proof over a verified stable public recommendation contract, durably preserving exact allocation identity/configuration and decision provenance. Recovery never re-selects an existing immutable Attempt; changed or unavailable Router state cannot rewrite it. No shadow routing, unsupported reservation claims, or Router execution-gateway dependency is introduced. Kernel retains unchanged Program intent/authority, Runtime execution/recovery, verification/acceptance/outcomes and Forge effects. An insufficient public execution-configuration contract stops #11 before implementation for an explicit owner decision.
A-F regressions are assigned to named child issues; #13 owns the B, D, and E external-operation regressions.

Constraints

Forgejo is canonical; implementation starts only from an owner-selected child issue.
Preserve adapters -> ports -> core; core remains usable without external products.
SQLite work verifies actual runtime capabilities and single-writer topology rather than imposing a universal version floor.
Use existing sandbox/container and credential mechanisms; do not build a custom sandbox.
No implementation changes, branches, PRs, merges, or modifications to main belong here.

Links

Governance: #1
K1: #4
K2A: #5
K2B: #13
K3: #6
K4: #7
K6: #8
K5: #9
K7: #10
K8: #11
Architecture: docs/architecture/target.md, docs/architecture/contracts.md, docs/architecture/successor.md, docs/architecture/threat-model.md
ADRs: docs/adr/0001-product-and-ownership.md through docs/adr/0006-context-and-outcomes.md

## Goal Coordinate the smallest deterministic Creatidy Kernel sequence that delivers a local durable Program with two dependent owner-approved engineering WorkUnits, deterministic recovery and acceptance semantics, replaceable resource/runtime/forge boundaries, and a reference flow that opens a Forgejo PR without automatic merge or deployment. ## Why A0 accepted the architecture and proved only the resource-boundary foundation. This record proves Kernel semantics before live integrations and keeps Forgejo, coding runtimes, and Scarcity Router outside core authority. ## Scope ### In scope - Coordinate implementation issues #4-#11 and #13. - Complete the first useful core path with FixedAllocator, local SQLite/artifacts, Forgejo, and one existing coding harness. - Treat #11 as an optional external-consumer ResourceAllocator proof after #10. The owner explicitly selected #11 for execution on 2026-09-27; the integration remains optional for Kernel users and does not transfer Program authority to Scarcity Router. Final dependency graph: ```text #4 K1 Domain and immutable intent -> #5 K2A Durable SQLite history and projections -> #13 K2B External operation journal, outbox, and artifact durability -> #6 K3 Synthetic execution, workspace, and authority enforcement -> #7 K4 Evidence-based verification, acceptance, remediation, and outcomes -> #9 K5 Forge port and Forgejo adapter -> #8 K6 Runtime port and Codex app-server adapter #9 + #8 -> #10 K7 CLI reference slice and restart/recovery proof -> #11 K8 optional external Scarcity Router ResourceAllocator proof ``` ### Out of scope REST, MCP, A2A, web UI, IDE integration, distributed workers/controllers, Temporal, Kubernetes, automatic merge, deployment, custom sandbox or secrets manager, custom policy DSL, adaptive allocation, Quota Governor, full Context Compiler, central telemetry, M5-B, Prefect, private migration, pilots, and releases. This is a coordination/roadmap record only. It does not authorize Program Execution Mode, autonomous task selection, or merge, main, release, deployment, or issue-creation authority. Each implementation issue must still be explicitly selected by the owner. ## Acceptance Criteria - [ ] Issues #4-#11 and #13 have executable scope, exclusions, acceptance evidence, exact numbered links, and architecture references. - [ ] #4-#7 and #13 prove deterministic and synthetic Kernel semantics without live Forgejo, Codex, Scarcity Router, M5-B, Prefect, or cloud services. - [ ] #10 drives a finite two-WorkUnit DAG through FixedAllocator, runtime, candidate retrieval, independent verification, acceptance, explicit successor input, restart/recovery, and Forgejo PR creation without automatic merge/deployment. - [ ] #11 repeats the accepted scenario as an optional external Scarcity Router ResourceAllocator proof over a verified stable public recommendation contract, durably preserving exact allocation identity/configuration and decision provenance. Recovery never re-selects an existing immutable Attempt; changed or unavailable Router state cannot rewrite it. No shadow routing, unsupported reservation claims, or Router execution-gateway dependency is introduced. Kernel retains unchanged Program intent/authority, Runtime execution/recovery, verification/acceptance/outcomes and Forge effects. An insufficient public execution-configuration contract stops #11 before implementation for an explicit owner decision. - [ ] A-F regressions are assigned to named child issues; #13 owns the B, D, and E external-operation regressions. ## Constraints - Forgejo is canonical; implementation starts only from an owner-selected child issue. - Preserve `adapters -> ports -> core`; core remains usable without external products. - SQLite work verifies actual runtime capabilities and single-writer topology rather than imposing a universal version floor. - Use existing sandbox/container and credential mechanisms; do not build a custom sandbox. - No implementation changes, branches, PRs, merges, or modifications to `main` belong here. ## Links - Governance: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/1 - K1: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/4 - K2A: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/5 - K2B: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/13 - K3: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/6 - K4: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/7 - K6: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/8 - K5: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/9 - K7: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/10 - K8: https://forgejo.creatidy.com/Creatidy/creatidy-kernel/issues/11 - Architecture: `docs/architecture/target.md`, `docs/architecture/contracts.md`, `docs/architecture/successor.md`, `docs/architecture/threat-model.md` - ADRs: `docs/adr/0001-product-and-ownership.md` through `docs/adr/0006-context-and-outcomes.md`

kilo referenced this issue

2026-09-22 16:49:34 +00:00

[K1] Domain and immutable intent #4

kilo referenced this issue

2026-09-22 16:49:35 +00:00

[K2A] Durable SQLite history and projections #5

kilo referenced this issue

2026-09-22 16:49:35 +00:00

[K3] Synthetic execution, workspace, and authority enforcement #6

kilo referenced this issue

2026-09-22 16:49:35 +00:00

[K4] Evidence-based verification, acceptance, remediation, and outcomes #7