Build #438 slice 6: coding map replaces the toy map #445

New issue

Closed

opened 2026-08-20 14:32:16 +00:00 by jared · 1 comment

jared commented

2026-08-20 14:32:16 +00:00

Owner

Parent

#438 (map-driven os-sdlc runner build spec). Design: docs/os-sdlc-rebuild/map-runner-design.md. Settled decisions + guard-manifest process: docs/os-sdlc-rebuild/interface-sketches.md.

Pre-build gate

First step before any code: run the seam-design pass per docs/os-sdlc-rebuild/seam-design-prompt.md to produce this slice's interface sketch. The sketch is the guard manifest: any class, public method, or signature change beyond it → stop, comment on this ticket, do not write it.

What to build

maps/coding.yaml expresses the real pipeline as data: the existing hardcoded test/lint gate chain ported into GateVerdict node data, AgentVerdict for the reviewer node, and gate fails routed to the terminal error node so shipped behavior is unchanged (ADR-0134). The toy map remains a test fixture.

Acceptance criteria

maps/coding.yaml validates and drives an implementation through the ported test/lint chain
Default coding-map behavior on a gate fail is identical to the current chain (terminal error node)
Reviewer node uses an agent-sourced verdict
Gate evidence lands in the runner-owned gate_results brief section, never inside handoffs

Blocked by

#444

Origin

Trigger: /to-tickets decomposition of #438 (2026-08-20, user-directed)
Improvised this session: none
Chain: DESIGN (docs/os-sdlc-rebuild/map-runner-design.md, docs/os-sdlc-rebuild/interface-sketches.md, ADR-0133/0134/0135)
Root candidate: #438

## Parent #438 (map-driven os-sdlc runner build spec). Design: docs/os-sdlc-rebuild/map-runner-design.md. Settled decisions + guard-manifest process: docs/os-sdlc-rebuild/interface-sketches.md. ## Pre-build gate First step before any code: run the seam-design pass per docs/os-sdlc-rebuild/seam-design-prompt.md to produce this slice's interface sketch. The sketch is the guard manifest: any class, public method, or signature change beyond it → stop, comment on this ticket, do not write it. ## What to build maps/coding.yaml expresses the real pipeline as data: the existing hardcoded test/lint gate chain ported into GateVerdict node data, AgentVerdict for the reviewer node, and gate fails routed to the terminal error node so shipped behavior is unchanged (ADR-0134). The toy map remains a test fixture. ## Acceptance criteria - [ ] maps/coding.yaml validates and drives an implementation through the ported test/lint chain - [ ] Default coding-map behavior on a gate fail is identical to the current chain (terminal error node) - [ ] Reviewer node uses an agent-sourced verdict - [ ] Gate evidence lands in the runner-owned gate_results brief section, never inside handoffs ## Blocked by - #444 ## Origin - Trigger: /to-tickets decomposition of #438 (2026-08-20, user-directed) - Improvised this session: none - Chain: DESIGN (docs/os-sdlc-rebuild/map-runner-design.md, docs/os-sdlc-rebuild/interface-sketches.md, ADR-0133/0134/0135) - Root candidate: #438

jared added the

plugin/os-sdlc

ready-for-agent

labels

2026-08-20 14:32:16 +00:00

jared referenced this issue

2026-08-20 14:32:30 +00:00

Build #438 slice 7: documentation sweep + plugin cache refresh #446

jared commented

2026-08-20 23:14:13 +00:00

Author

Owner

Resolution

Done: Slice 6 built TDD-first on build/438-slice-6, merged to main at 521f24c. Real coding map (maps/coding.yaml) ported the hardcoded test/lint gate chain as GateVerdict node data; AgentVerdict added for the reviewer node; gate fails route to terminal error node so shipped behavior is unchanged; UnknownGate validates unknown gate names in the map; schema 11 adds step_results.output for impl-owned output + gate_results brief section for runner-owned evidence (D31–D40, OQ3-A/OQ4-A approved in docs/os-sdlc-rebuild/interface-sketches.md).

Evidence: ruby -Ilib tests/all.rb: 850 runs, 0 failures (verified by orchestrator). rubocop zero offenses on all touched files, no disables. New maps/coding.yaml with test/lint gates + handler nodes; GateVerdict.extracted via schema 11 coding_gate payloads; UnknownGate guard on map.gate? lookups.

Follow-ups: Slice-7 seam pass (deferred to slice 7) owns the red_assert/test-writer gate shape (OQ2-B deferral). Follow-up ticket #451 filed for red_assert work. No new blockers; slice 7 is next (#446).

## Resolution **Done:** Slice 6 built TDD-first on build/438-slice-6, merged to main at 521f24c. Real coding map (maps/coding.yaml) ported the hardcoded test/lint gate chain as GateVerdict node data; AgentVerdict added for the reviewer node; gate fails route to terminal error node so shipped behavior is unchanged; UnknownGate validates unknown gate names in the map; schema 11 adds step_results.output for impl-owned output + gate_results brief section for runner-owned evidence (D31–D40, OQ3-A/OQ4-A approved in docs/os-sdlc-rebuild/interface-sketches.md). **Evidence:** ruby -Ilib tests/all.rb: 850 runs, 0 failures (verified by orchestrator). rubocop zero offenses on all touched files, no disables. New maps/coding.yaml with test/lint gates + handler nodes; GateVerdict.extracted via schema 11 coding_gate payloads; UnknownGate guard on map.gate? lookups. **Follow-ups:** Slice-7 seam pass (deferred to slice 7) owns the red_assert/test-writer gate shape (OQ2-B deferral). Follow-up ticket #451 filed for red_assert work. No new blockers; slice 7 is next (#446).

jared closed this issue

2026-08-20 23:14:15 +00:00

jared referenced this issue from a commit

2026-08-21 13:42:57 +00:00

feat(os-sdlc): real coding map and AgentVerdict extraction (#445)

jared referenced this issue from a commit

2026-08-21 13:42:57 +00:00

docs(os-sdlc): history row for #438 slice 6 (#445) and ADR-0137