diff --git a/README.md b/README.md
index a54c770..959a16d 100644
--- a/README.md
+++ b/README.md
@@ -7,7 +7,7 @@
 `intent → task semantics → catalog federation → live item evidence → asset plan → manifest`
 
 [![CI](https://github.com/GeoGeekLab/stac-scout/actions/workflows/ci.yml/badge.svg)](https://github.com/GeoGeekLab/stac-scout/actions/workflows/ci.yml)
-[![Python](https://img.shields.io/badge/Python-3.12%20%7C%203.13-3776AB?style=flat-square&logo=python&logoColor=white)](.github/workflows/ci.yml)
+[![Python](https://img.shields.io/badge/Python-3.12%20%7C%203.13%20%7C%203.14-3776AB?style=flat-square&logo=python&logoColor=white)](.github/workflows/ci.yml)
 [![License: MIT](https://img.shields.io/badge/license-MIT-2ea44f?style=flat-square)](LICENSE)
 [![STAC](https://img.shields.io/badge/STAC-evidence--first-111111?style=flat-square)](https://stacspec.org/)
 
@@ -129,17 +129,50 @@ confidence ≠ proof
 ## Install
 
 ```bash
-python -m pip install -e ".[dev]"
+python -m pip install stac-scout
 ```
 
-Python 3.12 or newer is required.
+Python 3.12 or newer is required. Python 3.12, 3.13, and 3.14 are tested in CI.
 
 For Planetary Computer access recipes:
 
 ```bash
-python -m pip install -e ".[planetary-computer]"
+python -m pip install "stac-scout[planetary-computer]"
+```
+
+## 60-second first success
+
+This first check is deterministic and does not call a remote catalog:
+
+```bash
+cat > request.json <<'JSON'
+{
+  "task": "assess wildfire impact",
+  "place": "Lahaina, Maui",
+  "datetime": {
+    "start": "2023-08-09T00:00:00Z",
+    "end": "2023-08-20T23:59:59Z"
+  }
+}
+JSON
+
+stac-scout advise request.json --task wildfire_impact
 ```
 
+The important part of the output is the decision contract, not prose:
+
+```json
+{
+  "data_type": "optical",
+  "required_measurements": ["nir", "swir22"],
+  "follow_up_requirements": ["comparison_windows"]
+}
+```
+
+Scout derives the optical/NIR/SWIR2 requirement but refuses to invent the missing comparison
+window. For a live Item-to-manifest walkthrough, see
+[`examples/wildfire-impact/`](examples/wildfire-impact/).
+
 ## Geo Task Intelligence
 
 ![Geo Task Intelligence workflow](docs/assets/geo-task-intelligence.svg)
@@ -501,8 +534,11 @@ stac-scout/
 │   ├── cases/
 │   ├── federation_cases/
 │   ├── task_cases/
+│   ├── regression_matrix.json
 │   ├── runner.py
 │   └── live.py
+├── examples/
+│   └── wildfire-impact/
 ├── docs/
 │   ├── ARCHITECTURE.md
 │   └── TASK_RULES.md
@@ -519,14 +555,14 @@ The deterministic evaluation corpus checks contracts that affect dataset decisio
 python evals/runner.py
 ```
 
-It covers:
+It covers request invariants, modality mismatches, cloud PASS/FAIL/UNKNOWN, required versus
+preferred measurements, DOI identity strength, multiple extents, antimeridian/polar AOIs,
+empty and >100 Item behavior, provider timeout/429/partial failure, categorical resampling,
+missing `file:size`, volume budgets, task/user conflicts, malformed provider metadata,
+cross-provider identity, task derivation, temporal strategy, and follow-up requirements.
 
-- request invariants;
-- cross-provider identity;
-- task-derived measurements;
-- task modality;
-- temporal strategy;
-- follow-up requirements.
+`evals/regression_matrix.json` binds the hardening acceptance scenarios to concrete pytest node
+IDs, so removing a required regression causes the eval runner to fail.
 
 Remote providers are intentionally kept out of ordinary CI.
 
@@ -548,7 +584,10 @@ remote catalog state
 
 ## Development
 
+Contributor installs are editable; normal users should use the PyPI command above.
+
 ```bash
+python -m pip install -e ".[dev]"
 ruff check .
 ruff format --check .
 mypy
@@ -557,7 +596,7 @@ python evals/runner.py
 python -m build
 ```
 
-CI runs on Python 3.12 and 3.13.
+CI runs on Python 3.12, 3.13, and 3.14. Python 3.15 prereleases are exercised as a non-blocking compatibility signal before the final 3.15 release.
 
 Coverage must stay at or above 90%.
 
diff --git a/examples/wildfire-impact/README.md b/examples/wildfire-impact/README.md
new file mode 100644
index 0000000..55aee8a
--- /dev/null
+++ b/examples/wildfire-impact/README.md
@@ -0,0 +1,98 @@
+# Wildfire impact: live evidence to replay
+
+This example exercises the complete STAC Scout decision path against a real STAC API. Remote
+catalog contents can change, so exact Item counts are observations rather than golden test values.
+The deterministic contracts behind the workflow are covered separately by CI and `evals/`.
+
+The AOI is a small polygon around Lahaina, Maui, and the Item window begins after the August 2023
+fire. A scientifically complete before/after impact study still requires an explicit comparison
+window; STAC Scout intentionally does not invent it.
+
+## 1. Confirm task requirements
+
+```bash
+stac-scout advise examples/wildfire-impact/request.json --task wildfire_impact
+```
+
+Expected semantics:
+
+```text
+data_type: optical
+required measurements: nir + swir22
+temporal strategy: before_after
+follow-up: comparison_windows
+```
+
+The checked-in request already carries the required modality and measurements so the remaining
+live steps use the same explicit hard constraints.
+
+## 2. Federate candidate discovery
+
+```bash
+stac-scout federate examples/wildfire-impact/request.json \
+  --max-workers 4 --overall-timeout 30
+```
+
+Inspect `candidates[]`, `duplicate_groups[]`, and typed `failures[]`. A provider failure does
+not erase candidates from other providers.
+
+## 3. Verify actual Items
+
+For a stable concrete path, use Element 84 Earth Search and Sentinel-2 L2A:
+
+```bash
+stac-scout verify examples/wildfire-impact/request.json \
+  --provider earth-search \
+  --collection sentinel-2-l2a \
+  --max-items 20
+```
+
+The result records observed Item IDs, timestamps, asset keys, AOI coverage, and warnings. If the
+provider returns exactly the cap, the observation is `limit_reached` rather than falsely called
+complete.
+
+## 4. Produce the asset plan, volume evidence, manifest, and recipe
+
+```bash
+stac-scout plan examples/wildfire-impact/request.json \
+  --provider earth-search \
+  --collection sentinel-2-l2a \
+  --max-items 20 \
+  --manifest scout.manifest.json \
+  --recipe load.py
+```
+
+Inspect:
+
+- `access_plan.asset_choices[]` for measurement-to-asset evidence;
+- each resampling decision and its evidence;
+- `estimated_bytes` when `file:size` metadata is available, otherwise explicit UNKNOWN;
+- output grid `utm` at 20 m;
+- search completeness and selected Item IDs in `scout.manifest.json`.
+
+The recipe is generated from the manifest; the manifest is the provenance record.
+
+## 5. Replay the observation
+
+```bash
+stac-scout replay scout.manifest.json
+```
+
+Replay reports complete/partial/inconclusive comparison status plus confirmed and unresolved
+missing/new Item IDs. A capped historical or current search is never promoted into false
+confirmed drift.
+
+## What this demo proves
+
+```text
+task semantics
+  → cross-provider candidates
+  → live Item evidence
+  → asset selection evidence
+  → resampling evidence
+  → volume estimate or UNKNOWN
+  → versioned manifest
+  → completeness-aware replay
+```
+
+It does not claim that one post-fire window alone is a complete wildfire impact analysis.
diff --git a/examples/wildfire-impact/request.json b/examples/wildfire-impact/request.json
new file mode 100644
index 0000000..99970d9
--- /dev/null
+++ b/examples/wildfire-impact/request.json
@@ -0,0 +1,24 @@
+{
+  "task": "assess wildfire impact around Lahaina after the August 2023 fire",
+  "geometry": {
+    "type": "Polygon",
+    "coordinates": [
+      [
+        [-156.72, 20.84],
+        [-156.64, 20.84],
+        [-156.64, 20.93],
+        [-156.72, 20.93],
+        [-156.72, 20.84]
+      ]
+    ]
+  },
+  "datetime": {
+    "start": "2023-08-09T00:00:00Z",
+    "end": "2023-08-20T23:59:59Z"
+  },
+  "data_type": "optical",
+  "required_measurements": ["nir", "swir22"],
+  "max_source_resolution_m": 20,
+  "target_crs": "utm",
+  "target_resolution_m": 20
+}