diff --git a/CHANGELOG.md b/CHANGELOG.md
index 87d781b..a31226c 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -15,17 +15,22 @@ The project uses a simple pre-release changelog during early development.
 - Positive and negative exchange-format fixtures.
 - Gate 0 review and closure records.
 - Gate 1 exchange-format freeze-candidate review.
+- Gate 1 exercise-identity collision review.
 - Stable exercise tracking modes.
 - Explicit load modes for no load, external resistance, and assistance.
 - Optional bounded session and exercise notes.
 - Extended body measurement list.
+- Executable local catalog reconciliation contract.
+- UUIDv4 generation policy for new Trainlog exercise and session IDs.
 
 ### Changed
 
 - Gate 0 project contract is complete.
-- Gate 1 is the active format-freeze gate.
+- Gate 1 remains active pending review #2 validation.
 - Actual repetition count may be zero for a failed attempt.
 - Planned exercises may contain zero actual sets.
-- Top-level exercise metadata must exactly match the session exercise references.
+- Top-level exercise metadata must exactly match session exercise references.
 - Assistance kilograms have distinct semantics from external resistance.
-- `TRAINLOG_FORMAT_V1` remains draft pending Gate 1 validation and mirrored review.
+- Different exercise IDs with equivalent normalized names are hard import conflicts.
+- Same exercise ID with incompatible tracking mode is a hard import conflict.
+- `TRAINLOG_FORMAT_V1` remains draft pending Gate 1 closure.
diff --git a/docs/android.md b/docs/android.md
index 947f1f2..8eba8e5 100644
--- a/docs/android.md
+++ b/docs/android.md
@@ -178,3 +178,30 @@ Initial Android versions do not need:
 - muscle classification;
 - distance/cardio metrics;
 - per-set rest measurement.
+## 13. Identifier generation
+
+When Android creates a new exercise, it generates:
+
+```text
+ex_<random UUID v4>
+```
+
+When Android creates a new session, it generates:
+
+```text
+se_<random UUID v4>
+```
+
+Display-name slugs must not be used as persistent identifiers.
+
+The visible exercise name remains independent from identity.
+
+## 14. Catalog conflict behavior
+
+Android must prevent duplicate normalized names inside its own local catalog.
+
+A valid Android export can still conflict with an independently edited TUI catalog.
+
+The TUI owns final reconciliation.
+
+Android must not assume that a matching display name means two different IDs may be silently merged.
diff --git a/docs/exchange_format.md b/docs/exchange_format.md
index 16916d6..5142458 100644
--- a/docs/exchange_format.md
+++ b/docs/exchange_format.md
@@ -480,3 +480,89 @@ Until then:
 ```text
 TRAINLOG_FORMAT_V1=DRAFT
 ```
+## 23. Generated identifier policy
+
+The wire format treats `exercise_id` and `session_id` as opaque identifiers satisfying their defined syntax.
+
+Official Trainlog implementations that create new identifiers must generate random UUID version 4 values.
+
+Generated exercise identifiers use:
+
+```text
+ex_<uuid-v4>
+```
+
+Generated session identifiers use:
+
+```text
+se_<uuid-v4>
+```
+
+Example:
+
+```text
+ex_550e8400-e29b-41d4-a716-446655440000
+se_550e8400-e29b-41d4-a716-446655440000
+```
+
+Importers accept any identifier valid under the v1 schema; they must not require that legacy or externally created IDs follow the UUID generation convention.
+
+The generation rule exists to prevent collisions when Android and TUI can both create exercises independently.
+
+## 24. Local catalog reconciliation
+
+Document validity and local importability are separate concepts.
+
+For each incoming exercise, the TUI reconciles against the local canonical catalog before the import transaction commits.
+
+### Same ID, same tracking mode
+
+Reuse the existing exercise.
+
+If the normalized display name differs, import is allowed but a non-fatal metadata warning is surfaced.
+
+The canonical local name is not silently changed.
+
+### Same ID, different tracking mode
+
+Reject the entire session import.
+
+A repetition identity and duration identity are semantically incompatible.
+
+### Different ID, equivalent normalized name
+
+Reject the entire session import as an identity conflict.
+
+Trainlog must not silently:
+
+- create a duplicate;
+- merge identities;
+- rewrite historical identifiers.
+
+The user must explicitly reconcile the conflict.
+
+### New ID, unique normalized name
+
+Create the exercise as part of the same database transaction as the session import.
+
+### Atomicity
+
+Any hard catalog conflict aborts the entire import.
+
+No partial session or exercise data may remain.
+
+## 25. Gate 1 executable import contract
+
+Before the production importer exists, catalog reconciliation is specified by:
+
+```text
+tests/contract/catalog-import-cases.json
+tools/validate_import_contract.py
+```
+
+Canonical validation requires both:
+
+```bash
+python tools/validate_json.py
+python tools/validate_import_contract.py
+```
diff --git a/docs/reviews/gate1_review_02.md b/docs/reviews/gate1_review_02.md
new file mode 100644
index 0000000..c9afb2f
--- /dev/null
+++ b/docs/reviews/gate1_review_02.md
@@ -0,0 +1,165 @@
+# Gate 1 Review #2 — Exercise Identity Collision Rules
+
+## Status
+
+```text
+GATE_1_REVIEW_01=IMPLEMENTED
+GATE_1_REVIEW_02=IMPLEMENTED
+GATE_1=VALIDATION_PENDING
+TRAINLOG_FORMAT_V1=DRAFT
+```
+
+## Purpose
+
+Review #1 defined the shape of Trainlog v1.
+
+Review #2 closes the remaining identity ambiguity between the Android exercise catalog and the canonical TUI catalog.
+
+This review is intentionally narrow.
+
+## G1-R2-01 — Generated identifiers must be globally collision-resistant
+
+The serialized format continues to treat `exercise_id` and `session_id` as opaque identifiers satisfying their documented syntax.
+
+Trainlog implementations that create new identifiers must use random UUID version 4 values.
+
+Recommended generated exercise identifier:
+
+```text
+ex_550e8400-e29b-41d4-a716-446655440000
+```
+
+Recommended generated session identifier:
+
+```text
+se_550e8400-e29b-41d4-a716-446655440000
+```
+
+The prefixes are part of the generated identifier.
+
+Importers must not require the UUID generation pattern for already-existing conforming v1 documents.
+
+This separates two concerns:
+
+- the wire format accepts the documented opaque identifier syntax;
+- official Trainlog implementations generate collision-resistant identifiers.
+
+## G1-R2-02 — Existing ID and compatible metadata
+
+If an incoming exercise has the same `exercise_id` and the same `tracking_mode` as an existing local exercise, the existing identity is reused.
+
+If the normalized display name is also equal:
+
+```text
+action = reuse
+```
+
+No warning is needed.
+
+## G1-R2-03 — Existing ID with a different display name
+
+If `exercise_id` and `tracking_mode` match but the display name differs:
+
+```text
+action = reuse_with_name_warning
+```
+
+The session may import.
+
+The local canonical exercise is not silently renamed.
+
+The UI must surface a non-fatal metadata warning.
+
+Deliberate rename/synchronization is a separate user action.
+
+## G1-R2-04 — Existing ID with incompatible tracking mode
+
+If the same `exercise_id` arrives with a different `tracking_mode`:
+
+```text
+action = reject_mode_conflict
+```
+
+The session import must not continue.
+
+The conflict cannot be repaired by silently changing history because `reps` and `duration` have different semantics.
+
+## G1-R2-05 — Different IDs with equivalent normalized names
+
+If an incoming exercise has a different `exercise_id` but its normalized name equals an existing local exercise name:
+
+```text
+action = reject_name_identity_conflict
+```
+
+The importer must not:
+
+- create a duplicate exercise;
+- silently merge the two identities;
+- silently rewrite the incoming session to the existing ID.
+
+The user must explicitly reconcile the identity conflict.
+
+This is the critical anti-duplicate rule for independently edited Android and TUI catalogs.
+
+## G1-R2-06 — New ID and unique normalized name
+
+If neither the identifier nor normalized name conflicts with the local catalog:
+
+```text
+action = create
+```
+
+The new exercise is inserted during the same import transaction as the session.
+
+## G1-R2-07 — Reconciliation is atomic
+
+Any hard catalog conflict rejects the entire session import transaction.
+
+The TUI must not partially:
+
+- create some new exercises;
+- insert the session;
+- insert sets.
+
+The database remains unchanged after a rejected import.
+
+## G1-R2-08 — Document validity versus local importability
+
+A Trainlog JSON document can be structurally and semantically valid on its own yet still be non-importable into a particular local database because of a catalog identity conflict.
+
+Therefore v1 has two distinct checks:
+
+```text
+document validation
+local catalog reconciliation
+```
+
+Gate 1 freezes both contracts.
+
+## Executable contract
+
+Canonical reconciliation cases live in:
+
+```text
+tests/contract/catalog-import-cases.json
+```
+
+Validate them with:
+
+```bash
+python tools/validate_import_contract.py
+```
+
+This validator is an executable specification used before the production C17 importer exists.
+
+## Freeze decision
+
+After both Gate 1 reviews are validated, committed, pushed, and mirrored, the closure step may set:
+
+```text
+GATE_1_REVIEW_01=PASS
+GATE_1_REVIEW_02=PASS
+GATE_1=PASS
+TRAINLOG_FORMAT_V1=FROZEN
+```
diff --git a/docs/roadmap.md b/docs/roadmap.md
index 4074d69..9241f39 100644
--- a/docs/roadmap.md
+++ b/docs/roadmap.md
@@ -11,16 +11,6 @@ GATE_0=PASS
 GATE_0_REVIEW_01=PASS
 ```
 
-Gate 0 established:
-
-- repository structure;
-- development contract;
-- architecture documentation;
-- coding-style rules;
-- exchange-format draft;
-- structural and semantic validation;
-- positive and negative fixture strategy.
-
 Reviewed hardening commit:
 
 ```text
@@ -35,43 +25,46 @@ ebd4316ed68c58598a471e567edf13455d00f92b
 
 ## Gate 1 — Exchange format v1 freeze
 
-Status: VALIDATION PENDING — REVIEW #1
+Status: VALIDATION PENDING — REVIEW #2
 
 Canonical state:
 
 ```text
 GATE_1_REVIEW_01=IMPLEMENTED
+GATE_1_REVIEW_02=IMPLEMENTED
 GATE_1=VALIDATION_PENDING
 TRAINLOG_FORMAT_V1=DRAFT
 ```
 
-Review #1 freezes the proposed model for:
+Review #1 defined:
 
-- exercise identity;
-- stable repetition/duration tracking mode;
-- session identity;
-- session ordering;
-- external/none/assistance load semantics;
-- planned versus actual sets;
-- zero-repetition failed attempts;
-- planned exercises with zero actual sets;
-- planned rest;
-- body weight;
-- final v1 body-measurement field list;
-- optional notes;
-- strict catalog completeness;
-- strict unknown-field behavior.
+- exercise identity and tracking mode;
+- session identity and ordering;
+- load semantics;
+- target versus actual work;
+- rest;
+- body weight and measurements;
+- notes;
+- strict document validation.
+
+Review #2 closes:
+
+- official UUIDv4 identifier generation;
+- Android/TUI catalog collision handling;
+- hard tracking-mode identity conflicts;
+- different-ID/same-name anti-duplicate conflicts;
+- atomic catalog reconciliation.
 
 Exit criteria:
 
-- all valid fixtures accepted;
-- all invalid fixtures rejected for the intended reason;
+- `python tools/validate_json.py` passes;
+- `python tools/validate_import_contract.py` passes;
+- `git diff --check` passes;
 - Android documentation aligned;
 - TUI documentation aligned;
-- schema and semantic validator aligned;
-- review commit pushed to Forgejo and GitHub;
+- both review commits pushed to Forgejo and GitHub;
 - mirrored review passes;
-- `TRAINLOG_FORMAT_V1=FROZEN`.
+- closure sets `TRAINLOG_FORMAT_V1=FROZEN`.
 
 ## Gate 2 — TUI persistence core
 
@@ -80,7 +73,9 @@ Deliverables:
 - Meson C17 project;
 - SQLite open/create;
 - schema versioning;
+- UUIDv4 identity generation;
 - exercise catalog;
+- atomic catalog reconciliation;
 - session import transaction;
 - idempotent import tests.
 
@@ -113,6 +108,7 @@ Exit criteria:
 Deliverables:
 
 - local exercise catalog;
+- UUIDv4 identity generation;
 - start/stop session timestamps;
 - target entry;
 - actual-set entry;
diff --git a/docs/tests.md b/docs/tests.md
index daf4297..fd27ea6 100644
--- a/docs/tests.md
+++ b/docs/tests.md
@@ -111,3 +111,28 @@ Before every meaningful push:
 4. run `git diff --check`;
 5. inspect `git status --short`;
 6. review documentation changes.
+## 10. Catalog reconciliation contract
+
+Before the C17 importer exists, Gate 1 defines local catalog merge behavior through an executable Python specification.
+
+Run:
+
+```bash
+python tools/validate_import_contract.py
+```
+
+Canonical cases cover:
+
+- exact existing exercise reuse;
+- same identity with renamed display text;
+- same identity with incompatible tracking mode;
+- different identities with equivalent normalized names;
+- new unique exercise creation.
+
+The full Gate 1 validation command is:
+
+```bash
+python tools/validate_json.py
+python tools/validate_import_contract.py
+git diff --check
+```
diff --git a/docs/tui.md b/docs/tui.md
index b071bcc..d3769c5 100644
--- a/docs/tui.md
+++ b/docs/tui.md
@@ -144,3 +144,47 @@ Numeric input is validated before persistent state is committed.
 Invalid input must never partially mutate a saved session.
 
 Imports use full validation before the database transaction commits.
+## 14. Catalog reconciliation during import
+
+Before creating any exercise or session rows, the TUI classifies incoming exercise metadata against the canonical local catalog.
+
+Rules:
+
+```text
+same ID + same mode + same normalized name
+    -> reuse
+
+same ID + same mode + different name
+    -> reuse + metadata warning
+
+same ID + different mode
+    -> reject entire import
+
+different ID + same normalized name
+    -> reject entire import
+
+new ID + unique normalized name
+    -> create inside import transaction
+```
+
+The TUI must never silently merge different exercise IDs merely because names match.
+
+The TUI must never create two identities with equivalent normalized display names.
+
+Any hard catalog conflict aborts the complete session import transaction.
+
+## 15. Generated identifiers
+
+When the TUI creates a new exercise directly, it generates:
+
+```text
+ex_<random UUID v4>
+```
+
+When the TUI creates a new session directly, it generates:
+
+```text
+se_<random UUID v4>
+```
+
+The database stores these identifiers as opaque stable text.
diff --git a/tests/contract/catalog-import-cases.json b/tests/contract/catalog-import-cases.json
new file mode 100644
index 0000000..67d8051
--- /dev/null
+++ b/tests/contract/catalog-import-cases.json
@@ -0,0 +1,86 @@
+[
+  {
+    "name": "same-id-same-name-same-mode",
+    "local": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Presse à cuisses",
+      "tracking_mode": "reps"
+    },
+    "incoming": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Presse à cuisses",
+      "tracking_mode": "reps"
+    },
+    "expected": "reuse"
+  },
+  {
+    "name": "same-id-renamed-same-mode",
+    "local": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Presse à cuisses",
+      "tracking_mode": "reps"
+    },
+    "incoming": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Presse jambes",
+      "tracking_mode": "reps"
+    },
+    "expected": "reuse_with_name_warning"
+  },
+  {
+    "name": "same-id-mode-conflict",
+    "local": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Gainage",
+      "tracking_mode": "reps"
+    },
+    "incoming": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Gainage",
+      "tracking_mode": "duration"
+    },
+    "expected": "reject_mode_conflict"
+  },
+  {
+    "name": "different-id-equivalent-name",
+    "local": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Presse à cuisses",
+      "tracking_mode": "reps"
+    },
+    "incoming": {
+      "exercise_id": "ex_22222222-2222-4222-8222-222222222222",
+      "name": "  PRESSE   À CUISSES ",
+      "tracking_mode": "reps"
+    },
+    "expected": "reject_name_identity_conflict"
+  },
+  {
+    "name": "different-id-equivalent-name-different-mode",
+    "local": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Gainage",
+      "tracking_mode": "reps"
+    },
+    "incoming": {
+      "exercise_id": "ex_22222222-2222-4222-8222-222222222222",
+      "name": " gainage ",
+      "tracking_mode": "duration"
+    },
+    "expected": "reject_name_identity_conflict"
+  },
+  {
+    "name": "new-id-unique-name",
+    "local": {
+      "exercise_id": "ex_11111111-1111-4111-8111-111111111111",
+      "name": "Presse à cuisses",
+      "tracking_mode": "reps"
+    },
+    "incoming": {
+      "exercise_id": "ex_22222222-2222-4222-8222-222222222222",
+      "name": "Gainage ventral",
+      "tracking_mode": "duration"
+    },
+    "expected": "create"
+  }
+]
diff --git a/tools/validate_import_contract.py b/tools/validate_import_contract.py
new file mode 100755
index 0000000..249c715
--- /dev/null
+++ b/tools/validate_import_contract.py
@@ -0,0 +1,71 @@
+#!/usr/bin/env python3
+"""Validate Trainlog v1 catalog reconciliation contract cases."""
+
+from __future__ import annotations
+
+import json
+import sys
+import unicodedata
+from pathlib import Path
+from typing import Any
+
+ROOT = Path(__file__).resolve().parents[1]
+CASES_PATH = ROOT / "tests" / "contract" / "catalog-import-cases.json"
+
+
+def normalize_exercise_name(name: str) -> str:
+    """Return the v1 normalized comparison form for an exercise name."""
+    nfc = unicodedata.normalize("NFC", name)
+    collapsed = " ".join(nfc.strip().split())
+    return collapsed.casefold()
+
+
+def classify(local: dict[str, Any], incoming: dict[str, Any]) -> str:
+    """Classify one incoming catalog entry against one existing local entry.
+
+    This function is executable specification for the Gate 1 import contract.
+    It does not mutate either object.
+    """
+    same_id = local["exercise_id"] == incoming["exercise_id"]
+    same_name = normalize_exercise_name(local["name"]) == normalize_exercise_name(
+        incoming["name"]
+    )
+    same_mode = local["tracking_mode"] == incoming["tracking_mode"]
+
+    if same_id:
+        if not same_mode:
+            return "reject_mode_conflict"
+        if same_name:
+            return "reuse"
+        return "reuse_with_name_warning"
+
+    if same_name:
+        return "reject_name_identity_conflict"
+
+    return "create"
+
+
+def main() -> int:
+    """Run every canonical catalog reconciliation case."""
+    with CASES_PATH.open("r", encoding="utf-8") as handle:
+        cases = json.load(handle)
+
+    failed = False
+
+    for case in cases:
+        actual = classify(case["local"], case["incoming"])
+        expected = case["expected"]
+
+        if actual != expected:
+            print(
+                f"FAIL {case['name']}: expected {expected!r}, got {actual!r}"
+            )
+            failed = True
+        else:
+            print(f"PASS {case['name']}: {actual}")
+
+    return 1 if failed else 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())