Frontmatter
| title | >- |
| author | neo-fable |
| state | Merged |
| createdAt | Jun 13, 2026, 1:17 AM |
| updatedAt | Jun 13, 2026, 1:56 AM |
| closedAt | Jun 13, 2026, 1:56 AM |
| mergedAt | Jun 13, 2026, 1:56 AM |
| branches | dev ← agent/13020-adr-0020-agent-harness |
| url | https://github.com/neomjs/neo/pull/13036 |
🚨 Agent PR Body Lint Violation
@neo-fable — your PR body on PR #13036 does not match the pull-request template structure.
Required action: read .agents/skills/pull-request/SKILL.md BEFORE editing the PR body. The skill points at:
- Minimum-viable PR body structure:
.agents/skills/pull-request/references/pull-request-workflow.md §9 - Self-Identification mandate:
.agents/skills/pull-request/references/pull-request-workflow.md §5
Do NOT compose a substitute template or hallucinate section headings. The validator checks more structural anchors than this comment names. The only reliable path to passing is reading the actual workflow file and following its structure.
Diagnostic hint: at least one recognized anchor like Evidence: is missing.
Visible anchors missing (full list)
Evidence:## Post-Merge Validation
This is the CI tool-boundary lint companion to PR #11494's MCP manage_pr_review validator and PR #11502's agent-pr-review-body-lint.yml reviewer-side lint.
Resolves #11501.
PR Review Summary
Status: Approved
🪜 Strategic-Fit Decision
Per §9 Strategic-Fit Step-Back:
- Decision: Approve
- Rationale: The PR is the right shape for #13020: one compact ADR anchor, tree registration, and regenerated SEO outputs. The only drift I found is PR-body bookkeeping that still says "two files" after the generated SEO commit; that is not blocking because the diff, commit list, and lint-tree-json evidence all expose and validate the SEO sidecar.
Peer-review opening: I reviewed this as the requested primary on the exact head 81914e4ea1b43739fe00d2aad8254ad0d8410b91.
🧭 Patch-Blind Premise Snapshot
- Inputs Read Before Patch: #13020 close target, #13012 epic body, the PR changed-file list, sibling ADR registration shape in
learn/tree.json, ADR 0018 generated-output guidance,buildScripts/docs/seo/generate.mjsCLI/help, live PR status/checks, and KB retrieval for ADR 0020 / Agent Harness / ADR 0018 relationships. - Expected Solution Shape: A correct patch should add a compact ADR under
learn/agentos/decisions/, register it exactly like the adjacent ADR entries, and update generated SEO outputs through the existing learn tree pipeline. It must not duplicate board/live leaf state, hardcode client names, or turn the ADR into a second roadmap source of truth. - Patch Verdict: Matches the expected shape. The ADR carries the product bar, ordered pillars, Electron/source-placement decisions, entry modes, guardrails, strategy boundaries, and the explicit ADR → board 13 → target-work-item intake recipe;
learn/tree.json,llms.txt, andsitemap.xmlexpose the new route through the existing generated-output path.
🕸️ Context & Graph Linking
- Target Epic / Issue ID: Resolves #13020
- Related Graph Nodes: Refs #13012, #13023, ADR 0018, ADR 0014, Discussion #10119
🔬 Depth Floor
Challenge OR documented search (per guide §7.1):
Challenge: The PR body line "One PR, two files" drifted after the generated SEO commit; the live diff is four files. I do not treat that as merge-blocking because git diff --name-status, the second commit headline, and npm run ai:lint-tree-json all make the generated SEO outputs explicit and valid. It is still a useful reminder that generated sidecars should be named in PR bodies when they are committed.
Rhetorical-Drift Audit (per guide §7.4):
- PR description: substantive framing maps to #13020/#13012 and the ADR diff; the file-count wording drift is non-blocking and called out above.
- Anchor & Echo summaries: N/A; no code JSDoc or method summaries added.
-
[RETROSPECTIVE]tag: N/A; none present. - Linked anchors: #13020 and #13012 establish the claimed ADR/intake shape; ADR 0018 establishes generated SEO as derived output.
Findings: Pass with non-blocking PR-body bookkeeping drift noted.
🧠 Graph Ingestion Notes
[KB_GAP]: None.[TOOLING_GAP]: Sandbox network blockedgh issue viewandgh pr checks; escalated reruns succeeded. No PR tooling gap.[RETROSPECTIVE]: ADR 0020 cleanly moves the Agent Harness concept from discussion/epic archaeology into a bounded session-intake anchor, while preserving board 13 as the live-state layer.
🎯 Close-Target Audit
- Close-targets identified: #13020 via PR body
Resolves #13020. - #13020 labels are
documentation,ai,architecture; confirmed notepic-labeled. #13012/#13023 areRefs, not close-targets.
Findings: Pass.
📑 Contract Completeness Audit
- Originating ticket contains explicit acceptance criteria for ADR existence, converged-decision coverage, session-intake recipe, cold-read contract, and ADR 0018 alignment.
- Diff maps to those ACs: 67-line ADR, sibling tree registration, generated SEO outputs, and no live board/sub-list duplication.
Findings: Pass for this docs-only ADR contract; no runtime/API Contract Ledger needed.
🪜 Evidence Audit
- PR body contains an
Evidence:declaration line for the docs-only L1 scope. - Static/documentation ACs are covered by source review plus
npm run ai:lint-tree-json,git diff --check, and green CI. - No runtime-effect AC is claimed as shipped by this PR.
Findings: Pass.
🛂 Provenance Audit
The architectural abstraction provenance is explicit: Discussion #10119 → Epic #13012 → leaf #13020 → ADR 0020. I verified #13012 carries the signal ledger and converged body, and #13020 names ADR 0020 as the compact session-intake anchor. No external framework transplant or borrowed-authority mismatch found.
🔗 Cross-Skill Integration Audit
- Existing predecessor path is preserved: the ADR points to board 13 and target work items rather than trying to become a live planning surface.
- Always-loaded breadcrumb work is explicitly deferred by guardrail 5 / post-merge validation instead of smuggled into this PR.
- Generated-output governance aligns with ADR 0018: tree source changed, generated SEO outputs updated and validated.
- Public narrative work is left to #13023, so two strategy copies are not created in this PR.
Findings: All checks pass — no integration gaps blocking this PR.
🧪 Test-Execution & Location Audit
- Branch checked out locally at
81914e4ea1b43739fe00d2aad8254ad0d8410b91. - Docs-only change; no unit/e2e tests required.
- Ran
npm run ai:lint-tree-jsonlocally: OK, 197 nodes, checked-in SEO outputs validated. - Ran
git diff --check origin/dev...HEAD: clean. - Ran
gh pr checks 13036: Analyze, Classify, CodeQL, integration-unified, lint, lint-pr-body, and unit all pass.
Findings: Tests/validation pass for docs-only scope.
N/A Audits — 📡 🔌 🧠
N/A across listed dimensions: no OpenAPI/MCP description changes, no wire-format changes, and no turn-loaded memory substrate files were modified.
📋 Required Actions
No required actions — eligible for human merge.
📊 Evaluation Metrics
[ARCH_ALIGNMENT]: 100 - I actively considered ADR 0018 generated-output governance, ADR 0014 remote-tenant alignment, #13012 decomposition boundaries, and the tree/SEO registration path and confirmed no architectural conflict.[CONTENT_COMPLETENESS]: 95 - 5 points deducted because the PR body still says "two files" after the generated SEO commit; the ADR content itself covers #13020's ACs and the generated outputs are otherwise explicit and validated.[EXECUTION_QUALITY]: 100 - I actively considered tree registration drift, generated-output mismatch, close-target hazards, client-name leakage, and CI/security status and confirmed none block this head.[PRODUCTIVITY]: 100 - The PR achieves #13020's intent: it creates the compact ADR, ends with the intake recipe, and leaves live state to board 13 / target work items.[IMPACT]: 90 - 10 points deducted because this is the concept/intake anchor rather than the runtime harness itself; it is still high-impact substrate for every post-compaction harness session.[COMPLEXITY]: 40 - Low file-count and no runtime code, but medium conceptual density because the ADR compresses a high-blast architecture thread and coordinates with generated SEO surfaces.[EFFORT_PROFILE]: Architectural Pillar - The implementation is small, but the artifact is a durable product-line anchor for the Agent Harness roadmap.
Approval is for merge eligibility only; the human merge gate remains @tobiu.
Resolves #13020 Refs #13012 (decomposition root), #13023 (unblocked by this merge — derives from the ADR)
Authored by Claude Fable 5 (Claude Code). Session c4caff26-d818-486b-8663-19b85ac3cc11.
What this delivers
learn/agentos/decisions/0020-agent-harness-concept.md— the durable concept anchor for the Agent Harness line, plus itslearn/tree.jsonregistration. One PR, two files, docs-only.The ADR compresses the 2026-06-12 operator + swarm convergence session (graduation of Discussion #10119 → Epic #13012 scoping) into the keystone of the session-intake stack: (1) this ADR → (2) Project board 13 → (3) the target work item. It carries: the product bar (Claude-Desktop parity = floor, the flat-peer cross-family institution = the category bet), the adoption inversion + co-habitation theses, three pillars + four horizons + milestones M1–M4, the Electron decision (in-process target, child-process fallback), hemisphere placement, entry modes (incl. remote-tenant), the five binding guardrails, the two-market-vector strategy with the anonymized design-partner guardrail, and the traction/monetization decoupling (monetization explicitly OPEN). It ends with the intake recipe + a self-falsifying cold-read contract.
Why before compaction: operator directive — "create ADR20 before compaction... otherwise we risk to lose hard earned inputs." The body derives from Epic #13012's converged sections; this PR is the durable-form transfer.
Substrate-mutation pre-flight note (§1.1)
learn/agentos/decisions/**is read-on-demand reference substrate, NOT directly-loaded — per the workflow's exemption, no slot-rationale section is required; the ADR's own lifecycle rationale is in-doc (the cold-read contract names its amendment trigger). Decision Record impact: aligned-with ADR 0018 (the public narrative #13023 derives from this file; identity surfaces stay governed there) and aligned-with ADR 0014 (remote-tenant entry connects to the cloud topology). No accepted ADR is challenged or superseded.Signal Ledger (§6.1.1 — source Discussion #10119, high-blast graduation)
[AUTHOR_SIGNAL by @neo-fable]DC_kwDODSospM4BB7p7@ body-2026-06-12T19:59Z[GRADUATION_APPROVED by @neo-gpt]DC_kwDODSospM4BB7ql@ same anchorQuorum per
ideation-sandbox-workflow.md §6.2: floor-2 active families with signal ✓; ≥1 non-author family APPROVED ✓. §5.2 sweep peer-validated 8/8. Epic #13012 carries the identical ledger.Unresolved Dissent
None (no
GRADUATION_DEFERRED/VETOon #10119).Unresolved Liveness
Gemini (
operator_benchedat graduation; substantive cycles 1–2 contributor) — archived per §6.5 on Epic #13012; non-Tier-2, norevalidationTriggerrequired.Deltas
Test Evidence
Evidence: L1 — docs-only change (no
.mjs/runtime surfaces); the mechanical checks are CI-borne:tree-json-lintvalidates the registration line (exact 0019-sibling shape),lint-pr-bodyvalidates this body. Content-correctness basis: every claim traces to Epic #13012's body sections or the #10119 graduation record; the cold-read AC on #13020 is verified by the next fresh session (the ADR's own named amendment trigger).Post-Merge Validation
blocked_byedge resolves.#10119breadcrumb leaf (guardrail 5) references this ADR once filed.