docs: sync post-P1 repository state

DOCUMENT-STATUS.md

@@ -1,6 +1,6 @@
 # LabPilot 文档状态
 
-> 当前状态：R0 仍为 V1 rebaseline authority root；Slice 1 已 Accepted；Slice 2 bounded MVP 已验证并集成到 `main`，production hardening deferred；Slice 3 Result Plane implementation 已完成。Competition Core Demo Runtime/E2E 已验收并冻结。P1 Result-Aware Experiment Iteration implementation completed and accepted; deterministic two-iteration acceptance + regression/full-CI evidence closed。
+> 当前状态：R0 仍为 V1 rebaseline authority root；Slice 1 已 Accepted；Slice 2 bounded MVP 已验证并集成到 `main`，production hardening deferred；Slice 3 Result Plane implementation 已完成。Competition Core Demo Runtime/E2E 已验收并冻结。P1 Result-Aware Experiment Iteration 已完成、验收并通过 PR #9 合并到 `main`（merge commit `03d1297e44307a37df1c9ed2b35e4dba60e00c87`）；当前没有活动中的 implementation tranche。
 
 | 文档 | 状态 | 用途 |
 |---|---|---|
@@ -19,7 +19,7 @@
 | Slice3 Detailed Implementation Plan | Approved after focused plan review | 冻结 Slice3 implementation 的 task/file/interface/TDD/acceptance 执行顺序；focused review C1–C5 为后续窄 authority |
 | Competition Core Demo closeout | **Accepted / Frozen** | Demo-specific Runtime/E2E closure：browser Golden Path + same immutable ExperimentVersion recovery replay + exact-authority CI；不构成 Slice 2 production acceptance |
 | P1 Result-Aware Experiment Iteration Design | **Approved / Current** | focused design review 已 APPROVE；P1-RESULT-001 CLOSED；作为当前 P1 设计 authority |
-| P1 Result-Aware Detailed Implementation Plan | **Approved / Current** | focused plan review APPROVE；P1-PLAN-001 CLOSED；正式授权按计划实施 P1-A/B |
+| P1 Result-Aware Detailed Implementation Plan | **Approved / Current** | focused plan review APPROVE；P1-PLAN-001 CLOSED；P1-A/B 已执行完成，保留为 accepted tranche 的 implementation authority |
 | Master Roadmap v0.2.1 | Historical frozen dependency | 仅用于历史 WP0 dependency verification |
 | WP0 Foundation v0.3.1 | Completed / Historical Accepted Execution Plan | WP0 已完成；不再标记 Ready for Execution |
 
@@ -118,6 +118,10 @@ Slice-specific detailed plans must record their exact planning baseline commit a
 - P1 deterministic two-iteration acceptance: **PASS**
 - P1 Task 8 focused regression: **PASS**
 - P1 Task 8 GitHub CI #312: **PASS**
+- P1 final closeout docs commit: `631b76190934ed7f6566f49998ae35b7fc274cef`
+- P1 final pre-merge GitHub CI #314: **PASS**
+- P1 PR #9: **MERGED**
+- P1 main integration merge commit: `03d1297e44307a37df1c9ed2b35e4dba60e00c87`
 
 Under Sequencing Amendment v0.3.1, the bounded Slice 2 MVP closeout is sufficient to unlock Slice 3. It does not retroactively complete the deferred Slice 2 production-hardening campaign or establish Slice 2 production acceptance.
 
@@ -127,12 +131,12 @@ The Git commit containing this status transition is the exact Slice3 implementat
 
 The competition workstream is closed. Its Runtime/E2E evidence and UI are frozen unless a concrete regression or demo-critical defect is demonstrated.
 
-P1 Result-Aware Experiment Iteration implementation is accepted. Repository integration is completed through the normal reviewed PR process; any further remaining-V1 tranche requires its own planning/authority gate.
+P1 Result-Aware Experiment Iteration implementation is accepted and merged into `main`. The repository is now in a post-P1 consolidation / maintenance state with **no active implementation tranche**. Any further remaining-V1 work requires a separate planning/authority gate before implementation.
 
 ```text
-implementation accepted
-→ reviewed PR integration
-→ separate planning/authority gate for any further tranche
+P1 accepted + merged
+→ consolidate / document / maintain
+→ separate planning/authority gate before any further tranche
 ```
 
 Competition closure still does not complete deferred Slice 2 production hardening or establish Slice 2 production acceptance.

README.md

@@ -8,7 +8,7 @@ LabPilot starts from a practical research problem: as experiments accumulate, th
 
 The goal is to make that research trajectory explicit and durable. LabPilot persists proposal revisions, approvals, immutable experiment versions, execution identities, physical attempts, artifacts, metrics, validations, and formal result selections so that an experiment can be traced from its research goal to its final result.
 
-The Agent participates on top of that lineage: it can inspect bounded repository context, incorporate prior tool results and human feedback, and help revise the next experiment proposal. The runtime and recovery mechanisms exist to make that assistance safe and reproducible rather than to turn the model into an unrestricted source of execution truth.
+The Agent participates on top of that lineage: it can inspect bounded repository context, incorporate prior tool results and human feedback, and—after a Batch closes—reason from an exact persisted result summary to help revise the next experiment proposal. The runtime and recovery mechanisms exist to make that assistance safe and reproducible rather than to turn the model into an unrestricted source of execution truth.
 
 ```text
 Research Goal
@@ -23,7 +23,9 @@ Research Goal
   -> Artifact / Metric / Validation
   -> Result Selection
   -> Batch Summary
-  -> Evidence for the next experiment iteration
+  -> Exact ResultContext
+  -> Follow-up AgentRun
+  -> Stop | next Proposal Revision in the same Experiment
 ```
 
 ## Why LabPilot
@@ -77,8 +79,12 @@ flowchart LR
     Metric --> Validation[Deterministic Validation]
     Validation --> Selection[Result Selection]
     Selection --> Summary[Batch Summary]
-    Summary --> Next[Next Experiment Iteration]
-    Next --> Goal
+    Summary --> ResultCtx[Exact ResultContext]
+    ResultCtx --> FollowAgent[Follow-up AgentRun]
+    FollowAgent --> FollowDecision{Stop / Human Input / Proposal}
+    FollowDecision --> FollowStop[Stop]
+    FollowDecision --> FollowHuman[Request Human Input]
+    FollowDecision --> Proposal
 ```
 
 ## Core capabilities
@@ -127,6 +133,8 @@ Tool Call | Request Human Input | Proposal | Stop
 
 This lets the Agent inspect evidence and help form the next proposal while keeping model authority explicit and every transition auditable.
 
+After a Batch summary is persisted, LabPilot can start a **Follow-up AgentRun** pinned to that exact `BatchSummaryRevision`. The model receives a bounded deterministic `ResultContext` containing the source experiment/version identities, Trial conditions, selected valid evidence, exclusions, warnings, and aggregates. A follow-up Proposal is created as the next `ProposalRevision` in the **same Experiment**, but it still must pass the existing deterministic verification and explicit human-approval path before a new `ExperimentVersion` exists. The Agent does not automatically approve the proposal or create/start the next Batch.
+
 ### 3. Reliable long-running physical execution
 
 Once an approved experiment crosses into real training, LabPilot has to preserve lineage even through crashes and uncertain external outcomes.
@@ -179,12 +187,13 @@ Runtime Events and SSE-backed read paths provide traceability and replay across
 | Launch intent before Runner launch | Gives recovery logic a durable fact to reconcile after crashes or uncertain external outcomes. |
 | Stable execution identity | Lets recovery refer to the same physical execution instead of accidentally creating a new one. |
 | Deterministic Result Validation | The formal scientific result should be reproducible and auditable instead of depending on a free-form LLM judgment. |
+| Exact result-context pinning | A Follow-up AgentRun reasons from one exact persisted BatchSummaryRevision rather than mutable latest/raw result state. |
 
 ## Runtime components
 
 - **FastAPI control plane** - experiment, proposal, execution, query, and event APIs.
 - **PostgreSQL** - authoritative experiment lineage, identities, command receipts, events, and result records.
-- **Bounded Agent Runtime** - repository-aware proposal iteration under explicit tool and budget constraints.
+- **Bounded Agent Runtime** - repository-aware and result-aware proposal iteration under explicit tool, budget, provenance, and approval constraints.
 - **Coordinator** - preparation, admission, execution control, recovery, cancellation, and reconciliation.
 - **Runner** - isolated physical-process boundary for real training workloads.
 - **Repository observation layer** - bounded, read-only repository access used by the Agent Harness.
@@ -247,6 +256,7 @@ The current core is not intended to be:
 - an unrestricted autonomous research agent;
 - a generic workflow engine;
 - a complete experiment-tracking UI or scientific notebook replacement;
+- an automatic self-running next-experiment loop that bypasses verification or human approval;
 - a claim of production-scale horizontal scheduling or global exactly-once semantics.
 
 Those are separate product and scaling problems. V1 first establishes a trustworthy experiment lineage and the invariants required for an AI Agent to participate in that lineage without making it less reliable.

docs/superpowers/evidence/2026-09-22-labpilot-p1-result-aware-iteration-closeout.md

@@ -2,7 +2,7 @@
 
 Date: 2026-09-22
 
-Status: Accepted
+Status: Accepted / Integrated
 
 ## 1. Authority
 
@@ -67,3 +67,12 @@ This closeout does NOT establish:
 - measured researcher productivity improvement.
 
 The product claim remains: bounded Agent-assisted Experiment Runtime / Harness.
+
+
+## 6. Repository integration
+
+- Final closeout documentation commit: `631b76190934ed7f6566f49998ae35b7fc274cef`.
+- Final pre-merge GitHub CI: #314 — SUCCESS, including production image/build-context hygiene.
+- PR #9 was merged through the reviewed PR path.
+- Main integration merge commit: `03d1297e44307a37df1c9ed2b35e4dba60e00c87`.
+- No follow-on implementation tranche is authorized by this closeout; further remaining-V1 work returns to a separate planning/authority gate.