Sync the Agent Systems Atlas profile example

README.md

@@ -167,9 +167,11 @@ generated contracts, Runtime Cards, and adoption receipts. It restores the
 previous control plane if any step fails.
 
 An empty corpus follows the same adoption contract. First perform bounded
-founding work to create real canonical owners and the residual-scan witness;
-then adopt a Profile revision that configures Corpus Planning before starting
-large-scale work. The exact sequence is documented in
+founding work to create real canonical owners and the residual-scan witness —
+one page may serve as both owner and witness when that is semantically
+natural, but pages are never merged only to save files. Then a second R09
+revision configures the Corpus Planning slot before large-scale work begins.
+The exact sequence is documented in
 [profiles/README.md](profiles/README.md#adoption-flow).
 
 ## Start Or Resume A Task

README.zh-CN.md

@@ -161,7 +161,8 @@ python3 Tools/apply_profile_adoption.py . --plan <plan>.yaml --apply
 采用 receipt 绑定在一起。任何一步失败，工具都会恢复之前的控制面。
 
 空语料库也使用同一份采用合同。先进行有界的 founding 工作，创建真实的
-canonical owner 和残留扫描见证；然后再次修订 Profile，配置 Corpus Planning，
+canonical owner 和残留扫描见证——语义自然时一页可以同时充当 owner 和见证，
+但绝不为了少建文件而强行合并；随后由第二次 R09 修订配置 Corpus Planning，
 之后才能开始大规模工作。完整顺序见
 [profiles/README.md](profiles/README.md#adoption-flow)。
 

profiles/examples/agent-atlas/README.md

@@ -2,17 +2,21 @@
 
 ## What This Example Is
 
-This directory is a non-normative, filled example of the Cambium profile interface for a real Chinese-first engineering knowledge corpus. It is not a template, default configuration, or adoption certificate. Because it lives under `profiles/examples/`, its manifest is intentionally not selectable in place.
+This directory is a non-normative snapshot of the real, selected Agent Systems
+Atlas Profile for a Chinese-first engineering knowledge corpus. It is not a
+template, default configuration, or adoption certificate. Because it lives
+under `profiles/examples/`, its manifest is intentionally not selectable in
+place.
 
 The normative interface is [profiles/README.md](../../README.md), and cross-domain rules remain in the kernel. A real adopter copies [profiles/_template/](../../_template/profile.md) to `profiles/<profile-id>/`, fills and validates that copy, and selects it through governance.
 
-The example does not contain or synchronize an adopter corpus, runtime state,
-or gate evidence. It remains a reference profile rather than an adoption
-certificate or proof of corpus-wide acceptance.
+The example does not contain the adopter corpus, runtime state, or gate
+evidence. It remains a reference profile rather than an adoption certificate
+or proof of corpus-wide acceptance.
 
 ## Directory Structure
 
-The directory contains 14 published files:
+The directory contains 17 published files:
 
 | Role | Count | Meaning |
 |---|---:|---|
@@ -24,6 +28,27 @@ The directory contains 14 published files:
 
 The scan configuration is not another slot or standard. This example has no separate Interview standards directory, supplemental Profile Read Set, profile-owned executable, or plugin.
 
+## Snapshot Provenance
+
+The 16 machine-active files in this directory were synchronized from the
+selected Profile in Agent Systems Atlas at commit
+`15df10eac89cafd381b145c48659c4a525f93f6d`. The source package was adopted
+under Cambium Standards 3.12.0.
+
+Only location-preserving adaptations are allowed in this example copy:
+
+- the Profile root changes from `profiles/agent-atlas/` to
+  `profiles/examples/agent-atlas/`;
+- relative links to the Profile interface and kernel execution-default
+  registry gain the additional `examples/` directory level;
+- the manifest title and introduction identify the package as a non-selectable
+  example.
+
+No Atlas rule, enum, trigger, role, judgment requirement, corpus path,
+structure declaration, scan predicate, or escalation threshold is abridged for
+the example. The orientation README is Cambium-owned and is not part of the
+selected Profile snapshot.
+
 ## Atlas-specific Bindings
 
 K11 owns the universal expression-layer floor. This profile supplies only the Atlas-specific answers: what an Interview artifact is, how it binds to canonical knowledge, how bilingual answers and readiness work, and what becomes a residual-content candidate.

profiles/examples/agent-atlas/escalation-policy.md

@@ -2,33 +2,32 @@
 
 Interface: [Escalation Policy slot](../../README.md#escalation-policy-slot)
 
-This corpus is built by long autonomous batch runs, so the executor reaches
-situations no batch gate judges: the plan itself changes shape, a batch stops
-moving, or an action cannot be undone. Each row below names one such
-situation, what counts as it happening, and who decides. The kernel trigger —
-explicit user authorization to modify the Standards or the selected profile —
-is owned by
+This corpus uses long autonomous batch runs, so the executor may reach
+situations that require a person before work can continue: the plan changes
+materially, a batch stops moving, or an action cannot be undone. The kernel
+trigger for modifying the Standards or the selected Profile is owned by
 [[kernel/K13 Task Runtime and Execution Control/17 Escalation Policy|K13/17]]
-and is deliberately not repeated here.
-
-The two magnitudes are this instance's numbers, recalculated at each
-maintenance boundary or every twenty closed batches. `replan_magnitude` is
-three times the largest change any ordinary replan has made to the Required
-record count, floored at 5%, and falls back to 30% while fewer than ten
-ordinary replans have run; an ordinary replan is one the user did not
-pre-review item by item. `batch_age` is ten times the p99 batch duration over
-a rolling twenty batches, clamped to between four and forty-eight hours, and
-falls back to forty-eight hours while fewer than twenty batches have closed.
-Both live in the Progress Ledger's run history, not in this file, because a
-value recomputed from the ledger cannot be restated here without going stale.
+and is not repeated here.
+
+Thresholds are derived from the Progress Ledger rather than copied into this
+Profile. At each maintenance boundary, or after another twenty batches close:
+
+- `replan_magnitude` is three times the largest Required-record-count change
+  made by an ordinary replan, with a 5% floor. Until ten ordinary replans have
+  completed, the threshold is 30%. An ordinary replan is one the user did not
+  pre-review item by item.
+- `batch_age` is ten times the rolling p99 duration of the latest twenty closed
+  batches, clamped to four through forty-eight hours. Until twenty batches
+  have closed, the threshold is forty-eight hours.
+
+The formulas are Profile policy; the current values remain runtime state.
 
 ## Escalation Triggers
 
 - Registration: Configured
 
 | Trigger ID | Condition that fires it | `machine-checkable` or `review-checkable` | Deciding Role ID reference | Resume condition |
 |---|---|---|---|---|
-| `quota_overrun` | A priority-quota target is exceeded with no recorded exemption | `machine-checkable` | `stopper` | The user grants an exemption, or the quota is brought back within target |
-| `replan_magnitude` | One replan changes the Required record count by more than the current `replan_magnitude` threshold | `machine-checkable` | `stopper` | The user approves the replanned scope |
-| `batch_stalled` | An `open` or `merge-ready` batch has been in that state longer than the current `batch_age` threshold | `machine-checkable` | `stopper` | The user rules on whether to finish, split, or abandon the batch |
+| `replan_magnitude` | One replan changes the Required record count by more than the ledger-derived `replan_magnitude` threshold. | `machine-checkable` | `stopper` | The user approves the replanned scope. |
+| `batch_stalled` | An `open` or `merge-ready` batch remains in that state longer than the ledger-derived `batch_age` threshold. | `machine-checkable` | `stopper` | The user decides whether to finish, split, or abandon the batch. |
 | `irreversible_operation` | The next action physically deletes a page, or rewrites history by any means other than a registered Amendment | `review-checkable` | `stopper` | The user authorizes that specific operation |

profiles/examples/agent-atlas/expression-layer.md

@@ -28,13 +28,16 @@ Interface: [Expression Layer Entry slot](../../README.md#expression-layer-entry-
 | Artifact role/type | `roadmap` |
 | Reader-facing display label | Interview Roadmap |
 | Entry point (target-corpus relative) | `Interview Preparation/Roadmaps/` |
-| Current pre-adoption target-corpus paths | `Agent Knowledge/Agent Interview Roadmap.md`; `Deep Learning Knowledge/DL Interview Roadmap.md`; `LLM Knowledge/LLM Interview Roadmap.md`; `Machine Learning Knowledge/ML Interview Roadmap.md` |
 | Profile-rule owner | `profiles/examples/agent-atlas/expression-layer.md#Interview Roadmap` |
 | Existing canonical dependency-map path | `Interview Preparation/Interview Overview.md` |
 | Testable regeneration/invalidation predicate | Re-evaluate an affected roadmap when its ordered topic set, prerequisite graph, or linked Interview Card inventory changes. |
 | Vocabulary Extensions readiness-field ID | `None` |
 | Routing/Gate Registry supplemental Gate ID | `None` |
 
+## Terminology Extraction Extension
+
+For a proper noun, satisfying the profile's [[profiles/examples/agent-atlas/expression-layer#Card Granularity|independent-Card predicate]] is an additional reason to consider a standalone Term Note under [[kernel/K05 Terminology/01 Terminology Extraction#Extraction Criteria|K05/01 Terminology Extraction]]. The Card still does not own the term's definition, and this trigger does not bypass the kernel's Do Not Extract or terminology-acceptance rules.
+
 ### Artifact: Interview Cheat Sheet
 
 | Property | Value |
@@ -43,17 +46,12 @@ Interface: [Expression Layer Entry slot](../../README.md#expression-layer-entry-
 | Artifact role/type | `cheat-sheet` |
 | Reader-facing display label | Interview Cheat Sheet |
 | Entry point (target-corpus relative) | `Interview Preparation/Cheat Sheets/` |
-| Current pre-adoption target-corpus paths | `Agent Knowledge/Agent Basics Cheat Sheet.md`; `LLM Knowledge/LLM Basics Cheat Sheet.md`; `Machine Learning Knowledge/Algorithms/ML Algorithms Checklist.md` |
 | Profile-rule owner | `profiles/examples/agent-atlas/expression-layer.md#Interview Cheat Sheet` |
 | Existing canonical dependency-map path | `Interview Preparation/Interview Overview.md` |
 | Testable regeneration/invalidation predicate | Re-evaluate an affected cheat sheet when a summarized canonical owner, linked card, or high-priority retrieval route changes. |
 | Vocabulary Extensions readiness-field ID | `None` |
 | Routing/Gate Registry supplemental Gate ID | `None` |
 
-## Terminology Extraction Extension
-
-For a proper noun, satisfying the profile's [[profiles/examples/agent-atlas/expression-layer#Card Granularity|independent-Card predicate]] is an additional reason to consider a standalone Term Note under [[kernel/K05 Terminology/01 Terminology Extraction#Extraction Criteria|K05/01 Terminology Extraction]]. The Card still does not own the term's definition, and this trigger does not bypass the kernel's Do Not Extract or terminology-acceptance rules.
-
 ## Interview Card Contract
 
 This is the stable profile-owned rule entry for Agent Systems Atlas Interview Cards. It owns the concrete Card behavior and organization described below. The cross-profile separation, evidence, status, linking, and migration floor remains owned by [[kernel/K11 Expression Layer Standard|K11 Expression Layer Standard]]; corpus placement remains owned by the profile's [[profiles/examples/agent-atlas/scope-and-architecture|Profile Scope]], and bilingual answer parity remains owned by its [[profiles/examples/agent-atlas/language-contract#Bilingual Answer Contract|Language Contract]].
@@ -192,16 +190,32 @@ A source note or public case cannot be presented as personal project evidence. Q
 
 ## Interview Roadmap
 
-An Interview Roadmap orders preparation by prerequisites, role relevance, priority, and verification milestones. The canonical entry for new Roadmaps is `Interview Preparation/Roadmaps/`; the pre-adoption paths listed in the artifact registration above remain migration inputs until moved. Roadmaps link to both canonical knowledge and applicable Interview Cards. They may name a practice or review checkpoint, but they do not contain the complete answers owned by Cards.
+An Interview Roadmap orders preparation by prerequisites, role relevance,
+priority, and verification milestones. Its canonical entry is
+`Interview Preparation/Roadmaps/`. Roadmaps link to both canonical knowledge
+and applicable Interview Cards. They may name a practice or review checkpoint,
+but they do not contain the complete answers owned by Cards.
 
 A roadmap checkbox reports the learner's progress on that roadmap only. It does not promote `authoring_status`, `interview_status`, evidence maturity, or corpus completion; the governing progress semantics are [[kernel/K11 Expression Layer/06 Sequence and Progress Semantics|K11/06 Sequence and Progress Semantics]].
 
 ## Interview Cheat Sheet
 
-An Interview Cheat Sheet is a compact retrieval aid whose canonical entry for new files is `Interview Preparation/Cheat Sheets/`; the pre-adoption paths listed in the artifact registration above remain migration inputs until moved. It may contain concise distinctions, decision cues, failure cues, formula reminders, and links to the canonical page and Interview Card. It does not own definitions, mechanisms, evidence, complete 30/90-second answers, or readiness decisions.
+An Interview Cheat Sheet is a compact retrieval aid whose canonical entry is
+`Interview Preparation/Cheat Sheets/`. It may contain concise distinctions,
+decision cues, failure cues, formula reminders, and links to the canonical
+page and Interview Card. It does not own definitions, mechanisms, evidence,
+complete 30/90-second answers, or readiness decisions.
 
 When a Cheat Sheet conflicts with a canonical owner or an accepted Card, it is stale and must be corrected from those owners; it is never used as the authority for resolving the conflict.
 
+`Interview Preparation/Cheat Sheets/ML Algorithms Cheat Sheet.md` derives from
+the canonical `Machine Learning Knowledge/Algorithms/ML Algorithms Checklist.md`
+and records that relation in its `canonical_bindings` frontmatter. The derived
+page keeps only interview retrieval cues, key distinctions, and links to
+canonical owners. When the canonical page's selection or acceptance criteria
+change, the derived page enters revalidation; the derived page's continued
+existence never proves synchronization.
+
 ## Related
 
 - [[kernel/Read Sets/R05 Expression Layer Read Set|R05 Expression Layer Read Set]]