Draft description, proposed IDs and legacy design migration (Stage 1, PR 7)

AGENTS.md

@@ -120,6 +120,10 @@ src/morphobricks_preprocessing/
   description_file.py   read_description(path): symlink refusal (R_FILE_SYMLINK), bytes -> Description
   run/description_file.py publish_description: lock, base SHA-256, strict re-read, transitions,
                     atomic write, append-only morphobricks.records.jsonl (Spec experiment-v1 4.6, 10)
+  draft_ids.py         proposed IDs: the folder label or raw header name only when the reader would
+                    accept it as written, else none (never repaired); draft.py the draft
+                    (IdSlot proposed/confirmed, open questions, unit_rows with condition_id,
+                    repeat_group, technical_repeat_id) and migrate_legacy_design
   image_type.py     image type (P4): FLUORESCENCE/RGB, review_image_type (suggestion from
                     TiffMetadata.rgb_samples and the confirmed roles, whether RGB is possible via the
                     brick's check_rgb), image_type_issues (rgb_not_possible)

CHANGELOG.md

@@ -100,6 +100,24 @@ are in [docs/plan.md](docs/plan.md) §12. Upgrading from v0.3: see the
   a changed image count alone is refused (new FOVs do not reopen a confirmation, section 10); the
   Spec's validator accepts it. Copying the records into the run manifest is a later pull request.
   Not used by any run, notebook or CLI command yet.
+- **Draft description, proposed IDs and legacy design migration (Stage 1, PR 7).** New
+  modules `draft_ids` and `draft`, and `design.parse_design_bytes` (the
+  former private `_parse_strict`, now public: the strict parse, no confirmation required). A folder label is proposed as a condition,
+  technical-repeat or biological-repeat ID only when the reader would accept it as written
+  (Spec 4.6a; the same rule for repeat IDs): `control`, `files` or `z` get no proposal, nothing is
+  repaired, derived or renamed. A raw header name is proposed as a channel ID only when it is a
+  valid ID as it stands (never trimmed or repaired; `R`, `G`, `B` never). A draft keeps IDs as
+  *proposed* until `confirm_id`, lists its open questions, and offers the added `condition_id`,
+  `repeat_group` and `technical_repeat_id` beside the unchanged source labels.
+  `migrate_legacy_design` reads `experiment_design.csv` strictly (`yes`/`no`/empty refused),
+  carries `repeat_group` as the proposed repeat ID, takes the latest `confirmed_on` of a group
+  with several dates and lists the original dates in a note (never the clock), maps 0, 1 or 2+
+  marked controls to no reference, one reference per condition or an open question, and refuses a
+  design that breaks a Spec rule with the reader's code. A legacy `repeat_group` that is not an
+  allowed repeat ID (`z`, `files`) is not refused: it migrates with no proposed ID and an open
+  question, so the user names the repeat (Spec 4.6a; owner decision of 7 November 2026). Where the Spec says "another allowed
+  ID" for a forbidden folder there is no proposal: no rule for making one was given. Nothing is
+  written and no run, notebook or CLI command uses it yet.
 - **Experiment-description model, canonical writer and section digests (Stage 1, PR 4).** New
   modules `description` (the `Description` model; `description_from_bytes` reads with the strict
   reader and raises `DescriptionError` with every issue of the first failing phase) and

pyproject.toml

@@ -178,6 +178,8 @@ baseline = "tools/code_rules_baseline.json"
 "morphobricks_preprocessing.design" = "io"
 "morphobricks_preprocessing.design_values" = "pure"
 "morphobricks_preprocessing.description_identity" = "pure"
+"morphobricks_preprocessing.draft_ids" = "pure"
+"morphobricks_preprocessing.draft" = "io"
 "morphobricks_preprocessing.physical_check" = "io"
 "morphobricks_preprocessing.description" = "pure"
 "morphobricks_preprocessing.description_write" = "pure"

src/morphobricks_preprocessing/design.py

@@ -608,7 +608,7 @@ def read_confirmed_design(
         )
     try:
         data = path.read_bytes()
-        design = _parse_strict(data, path.name)
+        design = parse_design_bytes(data, path.name)
     except (OSError, DesignFormatError) as exc:
         raise _unreadable(path, exc, rerun) from exc
     issues = design_issues(design)
@@ -740,8 +740,8 @@ def _unreadable(path: Path, exc: Exception, action: str) -> DesignCheckError:
     )
 
 
-def _parse_strict(data: bytes, name: str) -> ExperimentDesign:
-    """Parse a design as :func:`write_experiment_design` writes it, refusing anything else."""
+def parse_design_bytes(data: bytes, name: str = DESIGN_FILENAME) -> ExperimentDesign:
+    """Parse a design as :func:`write_experiment_design` writes it; confirmation not required."""
     if data.startswith(b"\xef\xbb\xbf"):
         raise DesignFormatError(
             f"{name} starts with a byte-order mark: it was saved by another program, not "