anatomize 2.0.0__tar.gz → 2.0.1__tar.gz

{anatomize-2.0.0 → anatomize-2.0.1}/CHANGELOG.md

@@ -2,6 +2,26 @@
 
 All notable changes to this project are documented in this file.
 
+## 2.0.1 - 2026-08-24
+
+### Changed
+
+- Explain the review model in reader language before introducing machine terms,
+  and describe imported evidence as saved results from familiar tools.
+- Make rendered dossiers and CLI help use task language while retaining stable
+  JSON field names for automation.
+
+### Fixed
+
+- Make session-store lock typing portable across Windows and POSIX type
+  environments.
+- Interpret imported POSIX and Windows absolute paths consistently on every
+  operating system, and read Unicode test fixtures explicitly as UTF-8.
+- Force the documentation verifier to install the built wheel when its isolated
+  environment can see an editable checkout.
+- Use an available `setup-uv` action and deploy the documentation site through
+  an enabled GitHub Pages environment.
+
 ## 2.0.0 - 2026-08-24
 
 ### Changed

{anatomize-2.0.0 → anatomize-2.0.1}/CITATION.cff

@@ -5,7 +5,7 @@ title: "Anatomize: evidence-first repository review for agentic development"
 authors:
   - family-names: Segal
     given-names: Bradley Max
-version: 2.0.0
+version: 2.0.1
 date-released: 2026-08-24
 license: MIT
 repository-code: "https://github.com/BradSegal/anatomize"

{anatomize-2.0.0 → anatomize-2.0.1}/CONTRIBUTING.md

@@ -21,7 +21,7 @@ package metadata and do not require uv.
   presentation in their owning packages; do not make a lower-level package
   call the review application.
 - Reuse shared path, identity, serialization, import, and error contracts.
-- Integrate specialist tools through an existing interchange before adding a
+- Integrate external analysis tools through an existing interchange before adding a
   vendor-specific adapter. Never add implicit tool execution or discovery.
 - Preserve provider provenance, conflicts, completeness, omissions, and
   limitations. Candidates must never encode an edit or deletion verdict.

{anatomize-2.0.0 → anatomize-2.0.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: anatomize
-Version: 2.0.0
+Version: 2.0.1
 Summary: Evidence-first repository review and lifecycle context for agents and humans
 Project-URL: Homepage, https://bradsegal.github.io/anatomize/
 Project-URL: Documentation, https://bradsegal.github.io/anatomize/
@@ -140,10 +140,11 @@ anatomize review change \
 Closure requires a fresh observation for every declared obligation. Anatomize
 accounts for the evidence but never fabricates a test, audit, or human approval.
 
-## Bring specialist evidence
+## Import results from tools you already use
 
-`review start` always builds the deterministic baseline itself. It can also
-import data artifacts that another tool has already produced:
+`review start` always maps the repository itself. It can also include saved
+results from tools you ran separately—for example, linter findings, test
+results, coverage, or a dependency inventory:
 
 ```bash
 anatomize review start . \
@@ -156,12 +157,12 @@ anatomize review start . \
   --output .anatomy/review/session.json
 ```
 
-Supported import kinds are `sarif`, `lsp`, `junit`, `coverage`, `mutation`,
+Supported file kinds are `sarif`, `lsp`, `junit`, `coverage`, `mutation`,
 `jscpd`, `snakemake`, `targets`, `renv`, `cyclonedx`, and `ro-crate`. Use
 `KIND@VERSION=PATH` when a native format needs an explicit version. A caller can
-also pass an already-normalised `ProviderEnvelope` with `--provider`. Anatomize
-does not discover plugins, execute providers, run repository code, or use the
-network during review.
+also pass a `ProviderEnvelope`: Anatomize's advanced, normalised format for
+results from another tool. Anatomize does not run any of these tools, run
+repository code, discover plugins, or use the network during review.
 
 ## Integrate an agent
 
@@ -190,26 +191,27 @@ as the CLI. It adds no second analysis path.
 - Artifacts bind repository identity, exact source state, producer evidence,
   configuration, and limitations.
 - Independent evidence is merged without last-writer-wins; explicit
-  cross-provider contradictions become conflict records.
+  contradictions between evidence sources become conflict records.
 - Sessions omit source text by default. `--include-source` is exact and
   opt-in.
 - Outputs are deterministic and human-readable; Markdown and plain text do not
   require ANSI or colour.
 - The library does not execute code, prove semantic equivalence, approve a
-  release, scan for secrets, or replace specialist compilers and test tools.
+  release, scan for secrets, or replace compilers, test runners, linters, or
+  security scanners.
 
 ## Documentation
 
 - [Documentation site](https://bradsegal.github.io/anatomize/): task-led paths
-  for reviewers, agent integrators, and provider authors.
+  for reviewers, agent integrators, and authors of external-tool integrations.
 - [Quickstart](https://bradsegal.github.io/anatomize/QUICKSTART/): the complete
   design-to-closure workflow.
 - [Lifecycle workflows](https://bradsegal.github.io/anatomize/WORKFLOWS/):
   design, audit, consolidation, implementation, and human hand-off.
 - [Interpret review results](https://bradsegal.github.io/anatomize/CONCEPTS/): status, evidence groups,
   omissions, conflicts, consolidation decisions, and closure.
-- [Provider artifacts](https://bradsegal.github.io/anatomize/PROVIDERS/): importing and authoring bounded
-  evidence.
+- [Import tool results](https://bradsegal.github.io/anatomize/PROVIDERS/): add
+  linter, test, coverage, dependency, and research-workflow output.
 - [Public reference](https://bradsegal.github.io/anatomize/REFERENCE/): CLI, Python namespaces, schemas, and
   supported formats.
 - [Troubleshooting](https://bradsegal.github.io/anatomize/TROUBLESHOOTING/): recovery and visible failure

{anatomize-2.0.0 → anatomize-2.0.1}/README.md

@@ -98,10 +98,11 @@ anatomize review change \
 Closure requires a fresh observation for every declared obligation. Anatomize
 accounts for the evidence but never fabricates a test, audit, or human approval.
 
-## Bring specialist evidence
+## Import results from tools you already use
 
-`review start` always builds the deterministic baseline itself. It can also
-import data artifacts that another tool has already produced:
+`review start` always maps the repository itself. It can also include saved
+results from tools you ran separately—for example, linter findings, test
+results, coverage, or a dependency inventory:
 
 ```bash
 anatomize review start . \
@@ -114,12 +115,12 @@ anatomize review start . \
   --output .anatomy/review/session.json
 ```
 
-Supported import kinds are `sarif`, `lsp`, `junit`, `coverage`, `mutation`,
+Supported file kinds are `sarif`, `lsp`, `junit`, `coverage`, `mutation`,
 `jscpd`, `snakemake`, `targets`, `renv`, `cyclonedx`, and `ro-crate`. Use
 `KIND@VERSION=PATH` when a native format needs an explicit version. A caller can
-also pass an already-normalised `ProviderEnvelope` with `--provider`. Anatomize
-does not discover plugins, execute providers, run repository code, or use the
-network during review.
+also pass a `ProviderEnvelope`: Anatomize's advanced, normalised format for
+results from another tool. Anatomize does not run any of these tools, run
+repository code, discover plugins, or use the network during review.
 
 ## Integrate an agent
 
@@ -148,26 +149,27 @@ as the CLI. It adds no second analysis path.
 - Artifacts bind repository identity, exact source state, producer evidence,
   configuration, and limitations.
 - Independent evidence is merged without last-writer-wins; explicit
-  cross-provider contradictions become conflict records.
+  contradictions between evidence sources become conflict records.
 - Sessions omit source text by default. `--include-source` is exact and
   opt-in.
 - Outputs are deterministic and human-readable; Markdown and plain text do not
   require ANSI or colour.
 - The library does not execute code, prove semantic equivalence, approve a
-  release, scan for secrets, or replace specialist compilers and test tools.
+  release, scan for secrets, or replace compilers, test runners, linters, or
+  security scanners.
 
 ## Documentation
 
 - [Documentation site](https://bradsegal.github.io/anatomize/): task-led paths
-  for reviewers, agent integrators, and provider authors.
+  for reviewers, agent integrators, and authors of external-tool integrations.
 - [Quickstart](https://bradsegal.github.io/anatomize/QUICKSTART/): the complete
   design-to-closure workflow.
 - [Lifecycle workflows](https://bradsegal.github.io/anatomize/WORKFLOWS/):
   design, audit, consolidation, implementation, and human hand-off.
 - [Interpret review results](https://bradsegal.github.io/anatomize/CONCEPTS/): status, evidence groups,
   omissions, conflicts, consolidation decisions, and closure.
-- [Provider artifacts](https://bradsegal.github.io/anatomize/PROVIDERS/): importing and authoring bounded
-  evidence.
+- [Import tool results](https://bradsegal.github.io/anatomize/PROVIDERS/): add
+  linter, test, coverage, dependency, and research-workflow output.
 - [Public reference](https://bradsegal.github.io/anatomize/REFERENCE/): CLI, Python namespaces, schemas, and
   supported formats.
 - [Troubleshooting](https://bradsegal.github.io/anatomize/TROUBLESHOOTING/): recovery and visible failure

{anatomize-2.0.0 → anatomize-2.0.1}/SECURITY.md

@@ -44,7 +44,7 @@ structure; they do not make embedded content safe to execute or obey.
 Anatomize is not a secret scanner, malware detector, vulnerability scanner,
 dependency resolver, package-signature verifier, authorization service,
 multi-user server, test runner, scientific-validity assessor, or release
-approval. Use the appropriate specialist tool and import its bounded result
+approval. Use the appropriate test, analysis, or security tool and import its saved result
 when it contributes to review.
 
 Before publishing an Anatomize artifact, inspect explicitly included source and

{anatomize-2.0.0 → anatomize-2.0.1}/docs/ARTIFACTS.md

@@ -10,8 +10,8 @@ validated object.
 | Review session | `review start` | Reuse one exact evidence state across questions | JSON |
 | Dossier exchange | `review dossier` / `expand` | Answer one bounded orientation, design, audit, localisation, implementation, change-review, or closure question | Markdown plus JSON when traceability matters |
 | Similarity artifact | `review similarity` | Discover conservative implementation, test, and documentation candidates | JSON |
-| Consolidation dossier | `review consolidate` | Adjudicate one candidate against differences and consequences | Markdown |
-| Decision overlay | `review overlay-create` | Preserve a consumer-owned disposition, owner, rationale, and review conditions | JSON or Markdown |
+| Consolidation dossier | `review consolidate` | Decide whether one possible duplicate should be merged, kept, or investigated further | Markdown |
+| Decision record (overlay) | `review overlay-create` | Preserve your disposition, owner, rationale, and review conditions separately from observed evidence | JSON or Markdown |
 | Overlay evaluation | `review overlay-check` | Determine whether relevant evidence changed after a decision | Markdown |
 | Implementation intent | `review intent` | Bind explicit obligations to the exact before-state | JSON |
 | Change dossier | `review change` | Compare canonical evidence across independent states | Markdown plus JSON |
@@ -24,7 +24,7 @@ Read in this order:
 
 1. **Status** — whether the evidence is complete, partial, or blocked.
 2. **Question and targets** — the exact decision boundary the dossier answers.
-3. **Proof-role groups** — why each fact matters, not merely where it was found.
+3. **Evidence groups** — why each fact matters, not merely where it was found.
 4. **Conflicts** — incompatible source-bound claims preserved without ranking.
 5. **Omissions and limitations** — absent evidence and what the methods cannot
    establish.

{anatomize-2.0.0 → anatomize-2.0.1}/docs/AUTOMATION.md

@@ -87,8 +87,9 @@ source inclusion.
 
 Start with orientation, then ask one targeted question. Prefer a stable file,
 symbol, test, documentation section, workflow, data item, or diagnostic over a
-repository-wide implementation request. Dossiers select required proof roles
-before optional context and preserve every reason an item was selected.
+repository-wide implementation request. Dossiers select the evidence purposes
+required by the task before optional context and preserve every reason an item
+was selected. The JSON field for an evidence purpose is `proof_role`.
 
 If a result is partial, choose an advertised action and preserve its opaque
 `action_id`. Actions support bounded budget increases, relationship traversal,

{anatomize-2.0.0 → anatomize-2.0.1}/docs/CONCEPTS.md

@@ -4,12 +4,32 @@ Anatomize outputs are designed to be read in a fixed order. Start with whether
 the question was answered, then inspect the selected evidence and anything that
 could change the decision.
 
+## A plain-language model
+
+You do not need to learn Anatomize's internal data model to use it. Seven terms
+appear in commands and output:
+
+| Term | Meaning |
+| --- | --- |
+| **Session** | A saved map of one exact checkout. Create it once with `review start`. |
+| **Dossier** | A short, task-focused selection from that map. It tells you what to inspect for one question. |
+| **Evidence source** | Anatomize itself or a tool whose saved results were imported. JSON calls this a `provider`. |
+| **Evidence group** | Items that serve the same purpose, such as definitions, callers, tests, or documentation. JSON calls this purpose a `proof_role`. |
+| **Decision record** | Your merge, keep, or postpone decision with its owner and rationale. The command name is `overlay-create`. |
+| **Implementation intent** | The requirements you record before editing. |
+| **Closure report** | The after-change check showing which recorded requirements have current supporting evidence. |
+
+An **artifact** is simply a saved input or output file, usually JSON, Markdown,
+or a tool's native report. A **source state** identifies the exact checkout the
+file describes. These names matter because a passing test report from yesterday
+must not be used as evidence for code changed today.
+
 ## Start with the status
 
 Every dossier reports one of three states:
 
-- `complete`: all required proof roles were available within the requested
-  boundary;
+- `complete`: all required evidence categories were available within the
+  requested scope;
 - `partial`: the result is useful, but a budget, scope, or provider left a
   material omission; or
 - `blocked`: the current session cannot support a defensible answer.
@@ -42,7 +62,7 @@ diagnostic rather than increasing every budget.
 
 ## Read the evidence groups
 
-Proof-role groups explain why an item was selected. Definitions and contracts
+Evidence groups explain why an item was selected. Definitions and contracts
 describe the target; consumers and relationships show what may be affected;
 tests, documentation, diagnostics, and runtime observations show where the
 change can be checked.
@@ -93,9 +113,10 @@ evidence and no declared unknown remains. Otherwise the report identifies an
 
 ## Choose the right output
 
-Use canonical JSON between tools and releases. Use Markdown for code review,
-design records, and other human hand-offs. Use plain text for terminals and
-logs, but do not parse it as an interchange format.
+Use canonical JSON—the stable, versioned JSON form—between tools and releases.
+Use Markdown for code review, design records, and other human hand-offs. Use
+plain text for terminals and logs, but do not parse it as an interchange
+format.
 
 You are ready to act when you can state the exact question, the evidence that
 supports the decision, the evidence that is absent or conflicting, and the

{anatomize-2.0.0 → anatomize-2.0.1}/docs/PROVIDERS.md

@@ -1,15 +1,24 @@
-# Provider artifacts and specialist integration
+# Import results from other tools
 
-A specialist result is useful to a review only when the consumer can identify
-what produced it, which repository state it describes, what it covered, and
-what it cannot establish. Anatomize therefore integrates tools through
-captured data rather than executing installed plugins. Acquisition remains
-under the caller's authority, and the resulting session can be reproduced in a
-different process or release.
+Anatomize maps source files, relationships, tests, documentation, and project
+configuration by itself. It does not run your linter, test runner, coverage
+tool, security scanner, dependency scanner, or research workflow. If you have
+already run one of those tools, you can add its saved result to the review.
 
-## Which import route should I use?
+This lets a dossier connect a finding or test result to the relevant code while
+keeping important limits visible. For example, an imported JUnit file describes
+only the tests selected in that run, and an old coverage file must not be used
+as evidence for a newer checkout.
 
-The normal route imports a supported native artifact:
+Anatomize calls each source of imported evidence a **provider**. Most users do
+not need to construct one: use `--artifact` with a supported result file. The
+`ProviderEnvelope` described later on this page is for authors building a new
+integration.
+
+## Import a supported result file
+
+Pass each file as `KIND=PATH`. Add `@VERSION` only when the format requires an
+explicit version:
 
 ```bash
 anatomize review start . \
@@ -21,7 +30,14 @@ anatomize review start . \
   --format json --output session.json
 ```
 
-The advanced route imports an already-normalised envelope:
+Anatomize checks that each result belongs to the same repository state before
+combining its evidence with the repository map.
+
+## Use the advanced integration format
+
+If you are writing an adapter for a result format that Anatomize does not
+support directly, normalise it to a `ProviderEnvelope` and pass that file with
+`--provider`:
 
 ```bash
 anatomize review start . \
@@ -29,10 +45,12 @@ anatomize review start . \
   --format json --output session.json
 ```
 
-Both routes are explicit. Every artifact is bound to the baseline repository
-and state before its evidence is merged.
+This route is intended for integration authors. It carries the tool name and
+version, exact checkout, configuration, scope, completeness, limitations, and
+normalised evidence in one validated file. Anatomize still does not execute the
+external tool.
 
-## Which artifact formats are maintained?
+## Supported result formats
 
 | Kind | Input | Evidence contributed |
 | --- | --- | --- |
@@ -57,9 +75,10 @@ A passing JUnit suite does not prove unselected tests passed. Coverage does not
 prove correctness. SARIF severity remains the producer's severity. A suggested
 fix remains untrusted input.
 
-## What must every provider envelope establish?
+## What the advanced provider format records
 
-The public `anatomize.providers` namespace supplies `ProviderEnvelope`,
+The rest of this page is API-level guidance for integration authors. The public
+`anatomize.providers` namespace supplies `ProviderEnvelope`,
 `ProviderBatchBuilder`, `ProviderScope`, `ProviderToolIdentity`,
 `build_provider_envelope`, parsing/writing helpers, evidence normalisation, and
 `run_provider_conformance`.
@@ -83,7 +102,7 @@ state is not approximately current evidence. Validation rejects the state
 mismatch so that a later closure report cannot silently use pre-change coverage
 as proof of the changed implementation.
 
-## How should a provider be authored?
+## Build a new integration
 
 A provider producer should:
 
@@ -107,7 +126,7 @@ contract are retained. When one explicitly conflicts and another supports or
 qualifies the target, evidence composition creates an unresolved
 `ConflictRecord`; it does not pick a winner.
 
-## When does a new adapter belong?
+## Decide whether a new adapter belongs
 
 Add a native adapter only when it contributes evidence that changes a named
 design, audit, localisation, consolidation, or closure decision. Reuse SARIF,
@@ -122,6 +141,6 @@ limitations. Anatomize intentionally has no provider execution broker or
 plugin discovery layer.
 
 Choose a native adapter when a maintained interchange already carries the
-needed evidence. Choose `ProviderEnvelope` when a specialist result needs a new
-normalisation layer. Add another native format only when neither route can
-represent evidence that materially changes a supported review decision.
+needed evidence. Choose `ProviderEnvelope` when an external tool's result needs
+a new normalisation layer. Add another native format only when neither route
+can represent evidence that materially changes a supported review decision.

{anatomize-2.0.0 → anatomize-2.0.1}/docs/QUICKSTART.md

@@ -36,9 +36,9 @@ anatomize review dossier .anatomy/review/before.json \
   --output .anatomy/review/orientation.md
 ```
 
-Read the status and proof-role groups first. They identify public surfaces,
-ownership, topology, tests, documentation, workflows, configuration, data, and
-available specialist evidence without dumping the repository.
+Read the status and evidence groups first. They identify public entry points,
+ownership, dependencies, tests, documentation, workflows, configuration, data,
+and any imported tool results without dumping the repository.
 
 ## 3. Localise the decision
 
@@ -90,7 +90,8 @@ anatomize review similarity .anatomy/review/before.json \
   --output .anatomy/review/similarity.json
 ```
 
-If a candidate exists, build its complete adjudication surface:
+If a candidate exists, collect the evidence needed to decide whether it should
+be merged:
 
 ```bash
 anatomize review consolidate \
@@ -103,7 +104,7 @@ anatomize review consolidate \
 
 Inspect the consolidation dossier before choosing a disposition. A helper that
 looks redundant may preserve different input handling or serve a distinct
-caller. Record the consumer-owned decision separately:
+caller. Record your decision separately from the observed similarity:
 
 ```bash
 anatomize review overlay-create \
@@ -154,8 +155,8 @@ anatomize review intent \
 ```
 
 Now edit. Run the repository's native tests, linters, documentation checks,
-security tools, and research validations. Anatomize imports their artifacts; it
-does not execute or impersonate them.
+security tools, and research validations. Anatomize can import their saved
+results; it does not execute or impersonate them.
 
 ## 6. Capture and compare the after-state
 

{anatomize-2.0.0 → anatomize-2.0.1}/docs/REFERENCE.md

@@ -27,8 +27,8 @@ source, budgets, and output paths are explicit arguments.
 | `review expand SESSION EXCHANGE ACTION` | Apply one source-bound advertised expansion without mutating the base exchange. |
 | `review similarity SESSION` | Project conservative implementation, test, and documentation candidates. |
 | `review change BEFORE AFTER` | Compare two exact states of one repository. |
-| `review consolidate EXCHANGE SIMILARITY CANDIDATE` | Gather the full adjudication surface for one candidate. |
-| `review overlay-create SESSION SIMILARITY CANDIDATE` | Record a portable consumer-owned disposition and rationale. |
+| `review consolidate EXCHANGE SIMILARITY CANDIDATE` | Gather the evidence needed to decide whether one candidate should be merged. |
+| `review overlay-create SESSION SIMILARITY CANDIDATE` | Record a portable decision, owner, and rationale without changing the evidence. |
 | `review overlay-check OVERLAY SESSION SIMILARITY CANDIDATE` | Evaluate decision currency against relevant current evidence. |
 | `review intent SESSION EXCHANGE OBLIGATIONS` | Bind explicit implementation obligations to the before-state. |
 | `review verify INTENT AFTER OBSERVATIONS` | Account for every obligation against fresh after-state evidence. |
@@ -100,7 +100,7 @@ Lower-level public ownership is:
 | `anatomize.index` | Deterministic baseline `RepositoryIndex` and `build_repository_index`. |
 | `anatomize.evidence` | Canonical entities, edges, contracts, observations, candidates, conflicts, completeness, omissions, aliases, lineage, merge, parse, and write. |
 | `anatomize.identity` | Portable identity keys, coordinate conversion, and cross-provider claim reconciliation. |
-| `anatomize.providers` | Data-only provider ABI, envelope construction, normalisation, conformance, and baseline projection. |
+| `anatomize.providers` | Advanced contract for normalising saved results from another analysis tool. |
 | `anatomize.sessions` | Portable session manifests/bundles and recoverable immutable stores. |
 | `anatomize.dossiers` | Targets, profiles, roles, requests, budgets, deterministic selection, slicing, cursors, and expansion. |
 | `anatomize.temporal` | Exact state manifests, evidence-wide comparisons, deltas, validation, and exact lineage. |

{anatomize-2.0.0 → anatomize-2.0.1}/docs/TROUBLESHOOTING.md

@@ -13,9 +13,9 @@ Only exact schemas advertised by this release are accepted:
 anatomize review capabilities --format json
 ```
 
-There is no migration layer. Regenerate a session from source or reacquire the
-specialist artifact with the current producer. Do not edit `schema_version`,
-IDs, or digests.
+There is no migration layer. Regenerate a session from source or rerun the tool
+that created the rejected result file. Do not edit `schema_version`, IDs, or
+digests.
 
 ## An artifact is stale
 
@@ -44,10 +44,11 @@ If no generation validates, remove or archive only that derived store and run
 `review start` again. The checkout and portable artifacts are authoritative;
 the store is disposable acceleration.
 
-## A provider is unavailable
+## Results from another tool are unavailable
 
-The baseline requires no optional tool. Run a specialist tool separately under
-your own authority, capture a supported artifact, and import it explicitly:
+Anatomize can still map the repository without optional tools. Run the linter,
+test runner, coverage tool, scanner, or workflow tool separately, save its
+result in a supported format, and import that file explicitly:
 
 ```bash
 anatomize review start . \
@@ -55,8 +56,9 @@ anatomize review start . \
   --format json --output session.json
 ```
 
-If the provider cannot run, retain an unavailable or partial status. Do not
-turn absence into a complete empty result.
+If the tool cannot run, retain an unavailable or partial status. Do not turn
+absence into a complete empty result. In JSON, the tool and its imported result
+are represented as a `provider`.
 
 ## A native artifact is rejected
 
@@ -90,9 +92,9 @@ anatomize review expand session.json exchange.json ACTION_ID \
   --format json --output expanded.json
 ```
 
-If a proof role is genuinely unavailable, acquire that evidence or record that
-the decision cannot yet close. Raising every budget is not a substitute for a
-precise target.
+If a required evidence category is genuinely unavailable, acquire that
+evidence or record that the decision cannot yet close. Raising every budget is
+not a substitute for a precise target.
 
 ## An expansion is rejected
 

{anatomize-2.0.0 → anatomize-2.0.1}/docs/WORKFLOWS.md

@@ -42,8 +42,8 @@ anatomize review dossier review/before.json \
 
 Read the dossier status before its items:
 
-- `complete` means every required proof role was available within the declared
-  boundary.
+- `complete` means every required evidence category was available within the
+  requested scope.
 - `partial` means useful evidence exists but a budget, provider, or scope left
   material omissions.
 - `blocked` means the question cannot be answered defensibly from the current
@@ -67,7 +67,7 @@ anatomize review dossier review/before.json src/anatomize/review/application.py
 
 The result groups the target definition, inbound and outbound relationships,
 consumers, tests, documentation, contracts, diagnostics, conflicts, and gaps
-by proof role. This is the context to inspect before editing, not a source dump.
+by purpose. This is the context to inspect before editing, not a source dump.
 
 ## Deduplicate and consolidate
 
@@ -88,9 +88,11 @@ The similarity artifact can contain implementation, test, and documentation
 candidates. If two path-normalisation helpers match, a consolidation dossier
 tests that resemblance against their behaviour, differences, consumers, tests,
 documentation, runtime evidence, contracts, history, conflicts, and unknowns.
-The consumer then records `merge`, `keep`, or `postpone` in a decision overlay
-with an owner and rationale. Imported jscpd regions can improve discovery, but
-the decision still depends on the complete adjudication surface.
+The reviewer then records `merge`, `keep`, or `postpone` with an owner and
+rationale. The JSON decision record is called an overlay because it adds
+judgement without changing the observed evidence. Imported jscpd regions can
+improve discovery, but the decision still depends on behaviour, callers,
+tests, documentation, contracts, history, conflicts, and remaining unknowns.
 
 ## Implement against explicit obligations
 
@@ -138,7 +140,7 @@ Keep canonical JSON for exact machine traceability. Export Markdown for the
 smallest set that materially helps a code reviewer:
 
 1. the targeted implementation or consolidation dossier;
-2. the decision overlay and rationale;
+2. the decision record (overlay) and rationale;
 3. the evidence-wide change dossier;
 4. relevant native test or analysis evidence; and
 5. the closure report.

{anatomize-2.0.0 → anatomize-2.0.1}/docs/generated/capabilities.md

@@ -3,7 +3,7 @@
 <!-- Generated by scripts/generate_documentation.py. Do not edit by hand. -->
 
 This page is generated from `anatomize.review.review_capabilities` in
-Anatomize `2.0.0`. Regenerate it whenever the public application
+Anatomize `2.0.1`. Regenerate it whenever the public application
 contract changes.
 
 | Contract | Value |

{anatomize-2.0.0 → anatomize-2.0.1}/docs/generated/python-api.md

@@ -2,7 +2,7 @@
 
 <!-- Generated by scripts/generate_documentation.py. Do not edit by hand. -->
 
-This inventory is generated from each public namespace's `__all__` under Anatomize `2.0.0`.
+This inventory is generated from each public namespace's `__all__` under Anatomize `2.0.1`.
 Names absent from `__all__` are implementation details.
 
 For most integrations, start with `anatomize.review.ReviewApplication` rather than lower-level builders.
@@ -247,7 +247,7 @@ Exported names: 29.
 
 ## `anatomize.dossiers`
 
-Task targets, proof roles, budgets, selection, and expansion.
+Task targets, evidence groups, limits, selection, and expansion.
 
 Exported names: 68.
 
@@ -273,7 +273,7 @@ Exported names: 68.
 | `DossierEngine` | class | Reusable deterministic query maps over one immutable dossier context. |
 | `DossierExpansion` | class | One consumer-requested expansion over an immutable base dossier. |
 | `DossierFilters` | class | Conjunctive restrictions applied without changing source facts. |
-| `DossierGroup` | class | Stable proof-role grouping in reader order. |
+| `DossierGroup` | class | Evidence items grouped by their purpose in reader order. |
 | `DossierItem` | class | One role-labelled evidence reference with every independent reason. |
 | `DossierLocator` | class | Structured portable target or evidence locator. |
 | `DossierObservation` | class | One independent provider observation retained without aggregation. |
@@ -528,9 +528,9 @@ Exported names: 37.
 | `canonical_sarif_bytes` | function | `(log: 'SarifLog') -> 'bytes'` — Serialize SARIF deterministically without reordering index-bearing arrays. |
 | `load_sarif_log` | function | `(path: 'Path', *, limits: 'SarifArtifactLimits' = SarifArtifactLimits(max_bytes=134217728, max_depth=96, max_values=1000000, max_string_bytes=2000000)) -> 'SarifLog'` — Load one SARIF log as untrusted bytes. |
 | `normalize_sarif_log` | function | `(log: 'SarifLog', *, binding: 'SarifArtifactBinding', expected_state: 'SourceStateRecord', expected_configuration_digest: 'str', baseline: 'RepositoryEvidence', sources: 'Mapping[str, str]', repository_uri_base_ids: 'Sequence[str]' = (), declared_contracts: 'Mapping[str, DeclaredDiagnosticContract] &#124; None' = None, invocation: 'InvocationAuthority &#124; None' = None) -> 'tuple[ProviderEnvelope, ...]'` — Normalize every SARIF run independently without comparing tool severities. |
-| `parse_sarif_log` | function | `(raw: 'bytes', *, limits: 'SarifArtifactLimits' = SarifArtifactLimits(max_bytes=134217728, max_depth=96, max_values=1000000, max_string_bytes=2000000)) -> 'SarifLog'` — Parse standard SARIF 2.1.0 without loading or executing a specialist tool. |
+| `parse_sarif_log` | function | `(raw: 'bytes', *, limits: 'SarifArtifactLimits' = SarifArtifactLimits(max_bytes=134217728, max_depth=96, max_values=1000000, max_string_bytes=2000000)) -> 'SarifLog'` — Parse a saved SARIF 2.1.0 report without loading or running its producer. |
 | `sarif_digest` | function | `(log: 'SarifLog') -> 'str'` — Return the canonical content identity of one parsed SARIF log. |
-| `unavailable_diagnostic_envelope` | function | `(*, tool_name: 'str', expected_state: 'SourceStateRecord', expected_configuration_digest: 'str', baseline: 'RepositoryEvidence', reason: 'str') -> 'ProviderEnvelope'` — Return explicit unavailable coverage without loading a specialist executable. |
+| `unavailable_diagnostic_envelope` | function | `(*, tool_name: 'str', expected_state: 'SourceStateRecord', expected_configuration_digest: 'str', baseline: 'RepositoryEvidence', reason: 'str') -> 'ProviderEnvelope'` — Record unavailable diagnostics without loading an external executable. |
 | `write_sarif_log` | function | `(log: 'SarifLog', path: 'Path') -> 'None'` — Atomically publish canonical SARIF while preserving array order. |
 
 ## `anatomize.research`

{anatomize-2.0.0 → anatomize-2.0.1}/docs/generated/schemas/dossier.schema.json

@@ -353,7 +353,7 @@
     },
     "DossierGroup": {
       "additionalProperties": false,
-      "description": "Stable proof-role grouping in reader order.",
+      "description": "Evidence items grouped by their purpose in reader order.",
       "properties": {
         "group_id": {
           "minLength": 1,

{anatomize-2.0.0 → anatomize-2.0.1}/docs/generated/schemas/dossier_exchange.schema.json

@@ -664,7 +664,7 @@
     },
     "DossierGroup": {
       "additionalProperties": false,
-      "description": "Stable proof-role grouping in reader order.",
+      "description": "Evidence items grouped by their purpose in reader order.",
       "properties": {
         "group_id": {
           "minLength": 1,