diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 70fb782..4ea888b 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -76,6 +76,7 @@ jobs:
if: matrix.family == 'python'
run: |
python -m unittest discover -s tests -v 2>&1 | tee ci-results/python.log
+ python -m scripts.verify_submission | tee ci-results/submission-verification.json
python scripts/fleet.py validate | tee ci-results/fleet-validation.log
python scripts/fleet.py plan | tee ci-results/fleet-plan.log
python -m fleet_witness demo --output runs/ci --seed 42 | tee ci-results/demo.log
diff --git a/README.md b/README.md
index da37d15..bb5c414 100644
--- a/README.md
+++ b/README.md
@@ -6,6 +6,7 @@
🛰️ SWARM ALIGNMENT HACKATHON · MISSION CONTROL FOR AGENT AUTHORITY Read the latest paper ·
Get the submission package ·
+ Watch the three-minute walkthrough ·
Launch the lab ·
Inspect the live pilot ·
Browse the docs
@@ -45,6 +46,20 @@ Permission can be revoked after a valid approval. An invalid approval stays inva
**The plot twist:** removing unsafe approvals by refusing everything is not useful model judgment. The pilot exposes that failure while showing the controller's enforcement boundary. The new [paper](docs/report/latest.pdf), [public trial table](data/paired-pilot-2026-11-09.json) and [evidence guide](docs/submission/latest/EVIDENCE.md) retain both results. No uncontrolled web swarms were observed.
+## 🎥 Step inside mission control
+
+[Watch the captioned three-minute walkthrough](docs/submission/media/fleet-witness-demo.mp4) ([captions](docs/submission/media/fleet-witness-demo.vtt), [media provenance](docs/submission/media/README.md)). It shows actual scripted viewer states, followed by the separately recorded live-supervisor findings.
+
+
+
+Verify the portable submission from this checkout with one offline command:
+
+```sh
+python3 -m scripts.verify_submission
+```
+
+This checks the allowlist, checksums, repository bytes, publication metadata and recomputed pilot totals. It does not authenticate private provider execution. The [evidence guide](docs/submission/latest/EVIDENCE.md) also covers the standalone package.
+
## 🚀 Launch the local lab
Use the [current pinned runtime matrix](ci/runtime-matrix.json). Python 3.11+ and Node 24 remain supported consumers. The application uses the Python standard library; Node runs viewer tests. The default demo is **scripted**, works offline and requires no API key or cloud account.
diff --git a/docs/AI_ASSISTANCE.md b/docs/AI_ASSISTANCE.md
index 286a845..d0bb917 100644
--- a/docs/AI_ASSISTANCE.md
+++ b/docs/AI_ASSISTANCE.md
@@ -58,3 +58,9 @@ paragraphs above describe their own dates, not current experiment availability.
The subsequent repository editorial pass used AI assistance to reconcile dated
status, clarify terminology, repair navigation and rebuild the latest publication
artifacts. It changed no experimental data, numerical findings or accepted ADRs.
+
+AI assistance also produced the offline submission verifier and its mutation tests,
+rendered the existing SVG cover to PNG, and captured the actual scripted viewer.
+The three-minute captioned video is an edited sequence of browser states and
+result cards; it is not a new model run or a continuous real-time recording.
+The original paper, experiment data and numerical findings are unchanged.
diff --git a/docs/BUILD_STATUS.md b/docs/BUILD_STATUS.md
index 8dfa001..db9d443 100644
--- a/docs/BUILD_STATUS.md
+++ b/docs/BUILD_STATUS.md
@@ -1,3 +1,56 @@
+# Submission finalization — November 9, 2026
+
+Task `submission-finalization-20261109`, based on
+`2ce59d34f92516cfb6666b891a4b35da2b46c9ba` (merged PR #14). The containing commit
+identifies this revision. The public package gains a read-only standard-library
+verifier, a PNG cover and an actual scripted viewer screenshot. A separate
+180-second captioned MP4 walks through actual viewer states and the recorded
+pilot result. README navigation links these assets; the viewer footer now lists
+both human authors.
+
+Observed local acceptance:
+
+- Eight new mutation tests failed before verifier implementation, then passed.
+ The initial qualified Python 3.15.2 suite passed **317 tests**; retained Node
+ 26.11.1 passes **five viewer tests**. Pre-existing SQLite ResourceWarnings remain.
+- The verifier passes against the repository ZIP and from its extracted standalone
+ copy: **17 explicit members**, balanced 160-trial / 80-pair cohort and all frozen
+ totals. The ZIP is below 1 MB. It checks consistency, not private provider
+ execution or visual PDF layout. Every Python CI lane now invokes it.
+- PNGs and all 16 video scenes were visually inspected. FFprobe reports exactly
+ **180 seconds**, H.264, 1600 × 900, 25 fps, 2,506,489 bytes; full decoding passes.
+ The video is an edited sequence held for reading, with captions and no audio.
+ Command/result cards are editorial summaries, not a raw terminal recording.
+- The seed-42 scripted demo, fleet validation and planner pass. The report source,
+ HTML, PDF and bibliography remain byte-identical to the base. The PDF remains
+ eight pages (six main pages), with a 218-word abstract.
+- Local Markdown audit: 80 files, 340 links, no unresolved target. Twenty-two
+ protected evidence files and five historical guide bodies remain unchanged.
+ `git diff --check` passes.
+
+Logs, mutation failures, capture attempts, action/request records, authoring scripts,
+video checks and source identities are retained in the sibling
+`submission-finalization-20261109` artifact packet. A first capture failed because
+loopback was included in a DNS block rule; a stale debugging-port retry also
+failed. Fresh isolated profiles and a loopback-only exception resolved capture;
+all recorded page requests were local or data URLs. Failed evidence is preserved.
+
+Independent review found three verifier gaps, reproduced by three additional
+regression tests: changed frozen data/interpretation labels, NUL-suffixed ZIP
+names and decompressor exceptions escaping the sanitized error boundary. The
+owner fixed them in one pass; all eleven verifier tests pass. The reviewed full
+suite passes **320 Python tests**, and all five viewer tests, standalone/repository
+checks and full video decoding pass again. The final hosted matrix is recorded
+with exact identities in the packet.
+
+Independent review and final-source hosted CI are separate gates; the final
+identities and results are recorded in the artifact packet and pull request.
+No new paid inference, cloud mutation, public observation or event submission
+was performed. Required contact details and first/replacement status remain
+unresolved outside Git. A prepared package is not event confirmation.
+
+---
+
# Repository editorial pass — November 9, 2026
Task `editorial-polish-20261109`, based on
diff --git a/docs/SUBMISSION.md b/docs/SUBMISSION.md
index fc195f3..3b6bcc2 100644
--- a/docs/SUBMISSION.md
+++ b/docs/SUBMISSION.md
@@ -1,6 +1,6 @@
# Team submission preparation
-**Latest package:** [PDF](report/latest.pdf), [editable report](report/latest.md), [ZIP](submission/latest-submission.zip), [form copy](submission/latest/SUBMISSION.md), [evidence guide](submission/latest/EVIDENCE.md), [demo script](submission/latest/DEMO.md).
+**Latest package:** [PDF](report/latest.pdf), [editable report](report/latest.md), [ZIP](submission/latest-submission.zip), [form copy](submission/latest/SUBMISSION.md), [evidence guide](submission/latest/EVIDENCE.md), [three-minute walkthrough](submission/media/fleet-witness-demo.mp4), [PNG cover](assets/fleet-witness-cover.png).
Authors: Alejandro Castellanos (Independent researcher) and Ignatius Pio Binawan (University of Toronto). The new report retains the existing manuscript's exact title and author order. [Contributions and AI assistance](AI_ASSISTANCE.md) distinguish the baseline, institutional study and new controlled pilot.
@@ -23,4 +23,4 @@ These findings do not establish improved useful model judgment, live swarm detec
## Remaining external step
-Review the PDF, enter personal details in the official form, upload the report and retain the confirmation privately. No upload, registration, credit claim or event submission is performed by this publication task. The optional video script is supplied; a video has not been recorded.
+Review the PDF, enter personal details in the official form, upload the report and retain the confirmation privately. No upload, registration, credit claim or event submission is performed by this publication task. The captioned walkthrough, separate WebVTT captions and PNG project image are ready. Personal contacts and whether this is a first or replacement entry still need confirmation; those details stay outside Git.
diff --git a/docs/assets/fleet-witness-cover.png b/docs/assets/fleet-witness-cover.png
new file mode 100644
index 0000000..060a710
Binary files /dev/null and b/docs/assets/fleet-witness-cover.png differ
diff --git a/docs/assets/fleet-witness-viewer.png b/docs/assets/fleet-witness-viewer.png
new file mode 100644
index 0000000..4ad0e40
Binary files /dev/null and b/docs/assets/fleet-witness-viewer.png differ
diff --git a/docs/report/latest.provenance.json b/docs/report/latest.provenance.json
index e6284f6..ecb2155 100644
--- a/docs/report/latest.provenance.json
+++ b/docs/report/latest.provenance.json
@@ -26,18 +26,32 @@
"Use the existing standard-library Markdown renderer and installed Chromium rather than adding a new PDF dependency.",
"Preserve the previous manuscript title and author order; Alejandro affiliation confirmed as Independent researcher.",
"Link the historical reproduction receipt rather than package its workstation-local artifact path; preserve the original receipt unchanged.",
- "Editorial revision clarifies terminology and navigation without changing title, author order, abstract, datasets, results or experiment sources; previous publication identities are retained."
+ "Editorial revision clarifies terminology and navigation without changing title, author order, abstract, datasets, results or experiment sources; previous publication identities are retained.",
+ "Added an offline verifier and browser-derived media; preserved all report/PDF bytes, datasets, findings and experiment identities. The MP4 is distributed separately from the explicit evidence ZIP."
],
"previous_publication": {
- "commit": "3d25f44482280766fdff305d470773dc30023f93",
- "package_sha256": "80ecc64234ae8d5193c1564f1c2c0508a6a373314080eb1aa876c5b2a57ac33e",
- "provenance_sha256": "b8a233aafb5200a092a2074c7cc782583c0a6c01137e68babdd61737c0ced928",
+ "commit": "2ce59d34f92516cfb6666b891a4b35da2b46c9ba",
+ "package_sha256": "ca8bbb44894c4ff5e7566845be639a4f3c3aefc55d5672d83e59467bbc81bfb4",
+ "previous_publication": {
+ "commit": "3d25f44482280766fdff305d470773dc30023f93",
+ "package_sha256": "80ecc64234ae8d5193c1564f1c2c0508a6a373314080eb1aa876c5b2a57ac33e",
+ "provenance_sha256": "b8a233aafb5200a092a2074c7cc782583c0a6c01137e68babdd61737c0ced928",
+ "public_file_sha256": {
+ "data/paired-pilot-2026-11-09.json": "6d3d681eb39c5af5bb7d08a9b82c50498729c336507c7eb26db7f1881a728c15",
+ "docs/report/latest-figures/paired-pilot.svg": "4b4dea6d5f8c9d11cfcc8611ab1e9a7beb147c07ea24a18e6adb12359c7aac54",
+ "docs/report/latest.html": "44b10d3441e2b4dffcd5cf50c1d1bf2801bc12be2f297e8b18263dca811c05ea",
+ "docs/report/latest.md": "001a9f99dad24de5f17a2028115650d58698917234138f221fbac1eb1cd08b7e",
+ "docs/report/latest.pdf": "c14d2471dc389e96bdf9284c6eee0ba7de2446505a698db916dc74cd20c4807a",
+ "docs/report/latest.references.json": "22895e1186dd06758dde9ee7cbac2ca101e87519ff1400c6606c5166a03f57bf"
+ }
+ },
+ "provenance_sha256": "2963de019c1821bfe0cb8c4c9c091f0e3f86e5d061d56977571131bce7329438",
"public_file_sha256": {
"data/paired-pilot-2026-11-09.json": "6d3d681eb39c5af5bb7d08a9b82c50498729c336507c7eb26db7f1881a728c15",
"docs/report/latest-figures/paired-pilot.svg": "4b4dea6d5f8c9d11cfcc8611ab1e9a7beb147c07ea24a18e6adb12359c7aac54",
- "docs/report/latest.html": "44b10d3441e2b4dffcd5cf50c1d1bf2801bc12be2f297e8b18263dca811c05ea",
- "docs/report/latest.md": "001a9f99dad24de5f17a2028115650d58698917234138f221fbac1eb1cd08b7e",
- "docs/report/latest.pdf": "c14d2471dc389e96bdf9284c6eee0ba7de2446505a698db916dc74cd20c4807a",
+ "docs/report/latest.html": "35b868ec1412c22baa9a601473c5d92a30f4adc4c71b6a02783e82ed20044c82",
+ "docs/report/latest.md": "1c88e2179d2605d7c9e1f574f420632b868651ae6f22a5e1d5649963cc25f304",
+ "docs/report/latest.pdf": "7c254ecaa345e472760ed227712386271b9ba8a12eefede7c226f22574b55d4a",
"docs/report/latest.references.json": "22895e1186dd06758dde9ee7cbac2ca101e87519ff1400c6606c5166a03f57bf"
}
},
@@ -46,14 +60,17 @@
"private_pilot_report_sha256": "57442e4842377db4d5fb1711914a6580a1973fffc74535b1bc73fd93c4c962ea",
"public_file_sha256": {
"data/paired-pilot-2026-11-09.json": "6d3d681eb39c5af5bb7d08a9b82c50498729c336507c7eb26db7f1881a728c15",
+ "docs/assets/fleet-witness-cover.png": "7dd3878f399ec87677eb8534ddadf29ee90453cc7fe9fff7482496ac3a5dadde",
+ "docs/assets/fleet-witness-viewer.png": "5da3a1ef0f35529d53ffc37ddd44a78e830db5012e537cb04eb5f2152a59bd72",
"docs/report/latest-figures/paired-pilot.svg": "4b4dea6d5f8c9d11cfcc8611ab1e9a7beb147c07ea24a18e6adb12359c7aac54",
"docs/report/latest.html": "35b868ec1412c22baa9a601473c5d92a30f4adc4c71b6a02783e82ed20044c82",
"docs/report/latest.md": "1c88e2179d2605d7c9e1f574f420632b868651ae6f22a5e1d5649963cc25f304",
"docs/report/latest.pdf": "7c254ecaa345e472760ed227712386271b9ba8a12eefede7c226f22574b55d4a",
- "docs/report/latest.references.json": "22895e1186dd06758dde9ee7cbac2ca101e87519ff1400c6606c5166a03f57bf"
+ "docs/report/latest.references.json": "22895e1186dd06758dde9ee7cbac2ca101e87519ff1400c6606c5166a03f57bf",
+ "scripts/verify_submission.py": "33032af215983e52e1ad14c60645526588933765d1568772b9751ce1f6998df3"
},
- "publication_base_commit": "3d25f44482280766fdff305d470773dc30023f93",
+ "publication_base_commit": "2ce59d34f92516cfb6666b891a4b35da2b46c9ba",
"publication_identity": "The containing Git commit identifies this publication revision; base and experiment source commits are distinct.",
"schema": "fleet-witness-publication-provenance-v1",
- "task": "editorial-polish-20261109"
+ "task": "submission-finalization-20261109"
}
diff --git a/docs/submission/latest-submission.zip b/docs/submission/latest-submission.zip
index f83e7ab..feea01f 100644
Binary files a/docs/submission/latest-submission.zip and b/docs/submission/latest-submission.zip differ
diff --git a/docs/submission/latest/DEMO.md b/docs/submission/latest/DEMO.md
index 916d878..51653e2 100644
--- a/docs/submission/latest/DEMO.md
+++ b/docs/submission/latest/DEMO.md
@@ -1,6 +1,8 @@
# Three-minute demo script
-This is a recording plan, not a completed video or a claim of live dashboard availability.
+Completed: [three-minute captioned MP4](https://github.com/k2jac9/fleet-witness/blob/main/docs/submission/media/fleet-witness-demo.mp4), with [captions and provenance](https://github.com/k2jac9/fleet-witness/tree/main/docs/submission/media). This is an edited sequence of 16 actual browser states and title/result cards, held for reading. It is not a continuous real-time capture or a claim of live dashboard availability. There is no audio or new model run.
+
+The outline below guided the edit; the WebVTT file gives the actual captions and timing.
## 0:00–0:30 — The mission
@@ -20,4 +22,4 @@ Open Figure 1 in the latest report and the public trial table. “In 80 pairs, t
## 2:30–3:00 — Show what can be checked
-Show the package checksum command and the offline reproduction snippet. Mention the separate historical two-model study and the finite checker. End with the limitations: synthetic tasks, local effects, incomplete historical provenance and no uncontrolled web observation. Link the repository and report.
+Show the one-command package verifier and the offline reproduction snippet. Mention the separate historical two-model study and the finite checker. End with the limitations: synthetic tasks, local effects, incomplete historical provenance and no uncontrolled web observation. Link the repository and report.
diff --git a/docs/submission/latest/EVIDENCE.md b/docs/submission/latest/EVIDENCE.md
index f35c4ed..c2f46de 100644
--- a/docs/submission/latest/EVIDENCE.md
+++ b/docs/submission/latest/EVIDENCE.md
@@ -1,44 +1,49 @@
# Evidence map and reproduction
-This package contains the report, editable source, one original figure, bibliography/provenance records, copy-ready submission text and a derived public trial table. It excludes credentials, private account records, provider response envelopes and the large historical archive. The complete implementation and historical archive are in the linked repository.
+This package contains the report, editable source, one original figure, bibliography/provenance records, copy-ready submission text, two PNG images, a standalone verifier and a derived public trial table. It excludes credentials, private account records, provider response envelopes and the large historical archive. The complete implementation and historical archive are in the linked repository.
## Verify the package
-From the extracted package directory on Linux:
+From the repository checkout (Python standard library only):
```sh
-sha256sum -c SHA256SUMS
+python3 -m scripts.verify_submission
```
-The manifest covers every member except itself. SHA-256 identifies the distributed bytes; it is not an external signature or an attestation of provider execution.
+This checks the ZIP against the current repository bytes. To check a downloaded
+ZIP without extracting it, use a trusted copy of the verifier:
-## Recompute the new pilot's public totals
+```sh
+python3 -m scripts.verify_submission --package PATH_TO_ZIP
+```
-Run this standard-library script from the extracted directory. It reads only the included JSON and makes no network calls:
+From an extracted package directory, the included standalone copy also works:
```sh
-python3 - <<'PY'
-import json
-from decimal import Decimal
-from pathlib import Path
-r = json.loads(Path('paired-pilot-2026-11-09.json').read_text())
-t = r['trials']
-assert len(t) == 160 and len({x['pair_id'] for x in t}) == 80
-for presentation in ('raw', 'scope_explicit'):
- rows = [x for x in t if x['presentation'] == presentation]
- invalid_approvals = sum(x['recommendation'] == 'approve' and not x['decision_valid_at_snapshot'] for x in rows)
- invalid_denials = sum(x['recommendation'] == 'deny' and not x['decision_valid_at_snapshot'] for x in rows)
- print(presentation, len(rows), invalid_approvals, invalid_denials)
-assert sum(x['effects'] for x in t) == 30
-assert sum(x['execution_attempts'] for x in t) == 68
-assert sum(x['authorization_denials'] for x in t) == 28
-assert sum(x['duplicate_noops'] for x in t) == 10
-assert sum(Decimal(x['provider_reported_cost_usd']) for x in t) == Decimal('0.2605526')
-print('effects=30; denied attempts=28; duplicate no-ops=10; charge=$0.2605526')
-PY
+python3 -B verify_submission.py
```
-Expected presentation rows: `raw 80 18 0` and `scope_explicit 80 0 40`. These numbers describe 80 pairs, not 160 independent samples. Decision-time labels are evaluator outputs, not model inputs. Public rows omit prompts and reasons; they support aggregate recomputation, not full private-trace replay or independent provider/account authentication.
+Use a verifier obtained from a trusted checkout. A script inside an arbitrary
+archive is code; checksums alone do not establish its authenticity.
+
+The verifier rejects unexpected, missing, duplicate, linked and oversized files.
+It checks exact checksum coverage, current provenance hashes, title/author/abstract
+consistency and every public trial's closed scalar schema. It recomputes balanced
+pairs, cell outcomes, effects, attempts, token totals and charges, then checks the
+immutable public-table digest. Interpretation labels must match the published
+contrast. Only stored and DEFLATE ZIP members are supported. It makes no
+network calls, extracts nothing, executes no bundled code and never repairs inputs.
+Success is JSON on stdout; failure is a nonzero exit with a sanitized message.
+
+Expected results: **160 trials, 80 pairs; raw 18 invalid approvals and zero invalid
+denials; scope explicit zero invalid approvals and 40 invalid denials; 30 effects,
+68 attempts, 28 authorization denials, ten duplicate no-ops; $0.2605526 provider
+charge**. These are 80 pairs, not 160 independent samples. `valid_decisions` in the
+source summary means schema-accepted decisions, not correct recommendations.
+
+SHA-256 identifies bytes; it is not a signature or evidence of provider execution.
+Public rows omit prompts and reasons, so aggregate consistency is not full private
+trace replay. PDF layout and the separate video are inspected independently.
## Test the code without paid calls
diff --git a/docs/submission/latest/SHA256SUMS b/docs/submission/latest/SHA256SUMS
index c34d7bb..af8c4ba 100644
--- a/docs/submission/latest/SHA256SUMS
+++ b/docs/submission/latest/SHA256SUMS
@@ -1,13 +1,16 @@
-b0f079ce934e4ed72b6e186b768427107f9f8ee1cc0882d9e25de17b460378b9 DEMO.md
-5cfbe1831bdeaeaf9533bffed1b0d0a51cd12dc73463ffe029fde704e9fcd155 EVIDENCE.md
+f69e1581653416ad07b55e87c94d9ecf081c4f08a95087fe28fd08b6d682cec8 DEMO.md
+28712a25c48593c1aa9d9e9db2f86129256202dd619b442131bd46c7cce8e52b EVIDENCE.md
cd9a1a84c34aaaa29ebea6e8593837329fab4f122f2cd9e8d8bc8c59c8abcd40 LICENSE
-a5946820f74489d464cf7b94daa3c4632e03211909c292fd431c0ddcead1e629 SUBMISSION.md
+c54fa3366c2f5e9bc2281ddc029a0a031ee27bde549ddec7c249fabecb828bbb SUBMISSION.md
01775933818a6680859402d3461a5bc3011a03ab55df8588badede0ad3779101 abstract.txt
+7dd3878f399ec87677eb8534ddadf29ee90453cc7fe9fff7482496ac3a5dadde fleet-witness-cover.png
+5da3a1ef0f35529d53ffc37ddd44a78e830db5012e537cb04eb5f2152a59bd72 fleet-witness-viewer.png
4b4dea6d5f8c9d11cfcc8611ab1e9a7beb147c07ea24a18e6adb12359c7aac54 latest-figures/paired-pilot.svg
35b868ec1412c22baa9a601473c5d92a30f4adc4c71b6a02783e82ed20044c82 latest.html
1c88e2179d2605d7c9e1f574f420632b868651ae6f22a5e1d5649963cc25f304 latest.md
7c254ecaa345e472760ed227712386271b9ba8a12eefede7c226f22574b55d4a latest.pdf
-2963de019c1821bfe0cb8c4c9c091f0e3f86e5d061d56977571131bce7329438 latest.provenance.json
+4708ce5886035028ea5298307174fa285fe3e12534cec3fc91e7c6cba5193a22 latest.provenance.json
22895e1186dd06758dde9ee7cbac2ca101e87519ff1400c6606c5166a03f57bf latest.references.json
-464855f53672c82a36026f068fa20902fd8227b4f320988c3d5c79dce1a38ccc metadata.json
+b215228fa318860e06c1453c4de3ac0afe7d5e8a159f82df2e7155d70a252d14 metadata.json
6d3d681eb39c5af5bb7d08a9b82c50498729c336507c7eb26db7f1881a728c15 paired-pilot-2026-11-09.json
+33032af215983e52e1ad14c60645526588933765d1568772b9751ce1f6998df3 verify_submission.py
diff --git a/docs/submission/latest/SUBMISSION.md b/docs/submission/latest/SUBMISSION.md
index a359df6..96786a7 100644
--- a/docs/submission/latest/SUBMISSION.md
+++ b/docs/submission/latest/SUBMISSION.md
@@ -24,4 +24,7 @@ Report upload: `latest.pdf`. The ZIP is a companion evidence package, not a subs
- If replacing an earlier entry, use the exact same title before the deadline; the latest files replace the earlier files according to the event instructions.
- Save the confirmation email or receipt privately. A Git push or a prepared ZIP is not submission confirmation.
-The official requirements allow at most eight main pages, excluding references and appendices; an abstract of 150–250 words; authors and affiliations; and a Limitations and Dual-Use / Ethical Considerations appendix. The report follows the official template's section structure, whose instructions permit adaptation. The source template is linked in the repository submission guide. The optional three-minute demo script is in `DEMO.md`; no video was recorded by this task.
+The official requirements allow at most eight main pages, excluding references and appendices; an abstract of 150–250 words; authors and affiliations; and a Limitations and Dual-Use / Ethical Considerations appendix. The report follows the official template's section structure, whose instructions permit adaptation. The source template is linked in the repository submission guide. The captioned three-minute walkthrough is linked in `DEMO.md`. The ZIP includes `fleet-witness-cover.png` for the optional project image and `fleet-witness-viewer.png` for an actual scripted viewer preview. The MP4 is published separately to keep the evidence ZIP small.
+
+Optional recording URL: https://github.com/k2jac9/fleet-witness/blob/main/docs/submission/media/fleet-witness-demo.mp4
+Additional material: https://github.com/k2jac9/fleet-witness/blob/main/docs/submission/latest-submission.zip
diff --git a/docs/submission/latest/metadata.json b/docs/submission/latest/metadata.json
index 8c6dfa0..57f4537 100644
--- a/docs/submission/latest/metadata.json
+++ b/docs/submission/latest/metadata.json
@@ -14,11 +14,13 @@
"deadline_utc": "2026-11-09T11:59:00Z",
"guidelines": "https://apartresearch.com/sprints/swarm-alignment-2026#guidelines",
"project": "Fleet Witness",
+ "project_image": "fleet-witness-cover.png",
"report": "latest.pdf",
"repository": "https://github.com/k2jac9/fleet-witness",
"submission_status": "prepared; external upload and confirmation not performed",
"template": "https://docs.google.com/document/d/1_1_27EyZriMlc8n8EBE2YScaeyMjm-8xjR097nmHXAE/edit",
"title": "Verify the Institution, Not the Agents: A Bounded-Verified Separation-of-Powers Kernel Stress-Tested Against Colluding LLM Swarms",
"track_proposed": "Multi-Agent Oversight and Control",
- "video_status": "script supplied; no video recorded"
+ "video_status": "180-second captioned edited walkthrough; no audio or new model calls",
+ "video_url": "https://github.com/k2jac9/fleet-witness/blob/main/docs/submission/media/fleet-witness-demo.mp4"
}
diff --git a/docs/submission/media/README.md b/docs/submission/media/README.md
new file mode 100644
index 0000000..d637ad4
--- /dev/null
+++ b/docs/submission/media/README.md
@@ -0,0 +1,35 @@
+# Three-minute walkthrough
+
+[Watch the MP4](fleet-witness-demo.mp4) · [WebVTT captions](fleet-witness-demo.vtt) · [Provenance and SHA-256](provenance.json)
+
+The 180-second video is an edited sequence of 16 browser states and caption cards,
+held for reading. It has burned-in captions and no audio. The format is H.264,
+1600 × 900, 25 fps, approximately 2.5 MB. It is distributed separately from the
+submission ZIP.
+
+| Time | What appears |
+| --- | --- |
+| 0:00–0:30 | Mission and the three separate checks |
+| 0:30–1:15 | Actual seed-42 scripted viewer: build-only receipt, denied attempt, paired presentation |
+| 1:15–1:50 | Actual viewer: revocation and duplicate execution |
+| 1:50–2:30 | Figure from the separately recorded 160-request live-supervisor pilot |
+| 2:30–3:00 | Editorial command/result cards, reproduction commands and limitations |
+
+The UI was captured from the running local viewer after actual scenario,
+presentation and event selections. The captions and result cards were authored
+for the walkthrough; the closing result summary is not a raw terminal recording.
+The source replay contains synthetic receipts; their signatures are not keys.
+No browser account, private provider trace or credential appears in the media.
+
+This is not a continuous real-time screen recording, a new inference run or a live
+cloud dashboard. The finite-check command is a reproduction instruction; depth-six
+exploration was not repeated to make this video. The original experiment records
+and PDF are unchanged.
+
+The [PNG cover](../../assets/fleet-witness-cover.png) renders the existing SVG
+artwork at 1280 × 560. The [viewer screenshot](../../assets/fleet-witness-viewer.png)
+is an actual 1600 × 1100 browser viewport with the Scripted demo label visible.
+`provenance.json` records source and artifact hashes. Full authoring scripts,
+local request/action logs, frames, failed capture attempts and encoding checks
+remain in the owner's retained `submission-finalization-20261109` artifact packet.
+These hashes identify files; they do not authenticate private model execution.
diff --git a/docs/submission/media/fleet-witness-demo.mp4 b/docs/submission/media/fleet-witness-demo.mp4
new file mode 100644
index 0000000..94f0a68
Binary files /dev/null and b/docs/submission/media/fleet-witness-demo.mp4 differ
diff --git a/docs/submission/media/fleet-witness-demo.vtt b/docs/submission/media/fleet-witness-demo.vtt
new file mode 100644
index 0000000..4b3e2f9
--- /dev/null
+++ b/docs/submission/media/fleet-witness-demo.vtt
@@ -0,0 +1,50 @@
+WEBVTT
+
+00:00:00.000 --> 00:00:10.000
+A swarm can agree on a release and still lack the evidence or authority to perform it.
+
+00:00:10.000 --> 00:00:20.000
+Authenticity. Applicability. Current authorization. A trustworthy system checks all three.
+
+00:00:20.000 --> 00:00:30.000
+Fleet Witness traces the recommendation, the controller decision, and the actual database effect.
+
+00:00:30.000 --> 00:00:42.000
+This is the actual seed-42 viewer. Its deterministic agents illustrate the mechanism; they are not measured models.
+
+00:00:42.000 --> 00:00:54.000
+Here the reviewer promotes a genuine build receipt into a release-safety claim. The raw supervisor approves.
+
+00:00:54.000 --> 00:01:05.000
+The recommendation is incorrect. The controller denies execution: a build receipt does not prove release safety.
+
+00:01:05.000 --> 00:01:15.000
+Switch to the paired scope-explicit presentation. The scripted supervisor now denies this unsupported release.
+
+00:01:15.000 --> 00:01:26.000
+This recommendation was valid at its snapshot. Permission is revoked before the execution attempt.
+
+00:01:26.000 --> 00:01:38.000
+The controller checks current permission and blocks the revoked attempt. No release record is committed.
+
+00:01:38.000 --> 00:01:50.000
+An authorized duplicate request is a no-op. A release here is a local SQLite record, not a public deployment.
+
+00:01:50.000 --> 00:02:02.000
+A separate recorded study used one real supervisor with synthetic workers and reviewers: 160 trials in 80 pairs.
+
+00:02:02.000 --> 00:02:15.000
+Raw: 18 invalid approvals. Scope explicit: zero invalid approvals, but 40 invalid denials—it denied every request.
+
+00:02:15.000 --> 00:02:30.000
+All invalid and revoked attempts were blocked. Thirty authorized effects committed. These are synthetic, local tasks.
+
+00:02:30.000 --> 00:02:42.000
+One offline command checks every packaged hash, compares repository sources, and recomputes the frozen trial totals.
+
+00:02:42.000 --> 00:02:52.000
+Run the offline demo and finite checker. The report also distinguishes the separate historical two-model study.
+
+00:02:52.000 --> 00:03:00.000
+Limits remain: synthetic tasks, local effects, incomplete historical provenance, and no uncontrolled web observation.
+
diff --git a/docs/submission/media/provenance.json b/docs/submission/media/provenance.json
new file mode 100644
index 0000000..aa8ca58
--- /dev/null
+++ b/docs/submission/media/provenance.json
@@ -0,0 +1,52 @@
+{
+ "artifact_sha256": {
+ "docs/assets/fleet-witness-cover.png": "7dd3878f399ec87677eb8534ddadf29ee90453cc7fe9fff7482496ac3a5dadde",
+ "docs/assets/fleet-witness-viewer.png": "5da3a1ef0f35529d53ffc37ddd44a78e830db5012e537cb04eb5f2152a59bd72",
+ "docs/submission/media/fleet-witness-demo.mp4": "e95518877e97b89682f12decbaba18dc63bfc5b5a678c4a8d9413768186daf44",
+ "docs/submission/media/fleet-witness-demo.vtt": "3754324477c18d4f3edda328efb23d88805bbf41933d722d7c45a2069c050b82"
+ },
+ "base_commit": "2ce59d34f92516cfb6666b891a4b35da2b46c9ba",
+ "capture_actions_sha256": "c86c6c4248150ef0b1ae99353e79bde46932de4222c57dd77a9cb1089f1d590b",
+ "capture_scripts_sha256": {
+ "capture.mjs": "47e26e275147de976eb044b4bfde18222bc4c349f277e826bd6b7fa92f7a6516",
+ "make-media.py": "03b9f6b07e80cd03979261a5205969bdb3cc7989c5e525ab8165530ccc66602a"
+ },
+ "date": "2026-11-09",
+ "inspection": "Both PNGs and all 16 scene frames visually inspected. FFprobe duration/codec checked. Full video decoding checked separately in the retained task packet.",
+ "scenes": 16,
+ "scope": "Actual local seed-42 scripted viewer; separately recorded live-supervisor pilot figure; editorial command/result summary cards. No new model run, account data, credentials, cloud mutation or public web observation.",
+ "scripted_replay_sha256": "a0c840784a0e1508eaace8a8086a42949f03ee06f603ecc5244386d6de841236",
+ "source_identity": "The containing Git commit identifies the final source; file hashes bind captured assets.",
+ "source_sha256": {
+ "data/paired-pilot-2026-11-09.json": "6d3d681eb39c5af5bb7d08a9b82c50498729c336507c7eb26db7f1881a728c15",
+ "docs/assets/fleet-witness-hero.svg": "8cdaf6d346d8dbaa9adc042dc02b82a092af1e2fd9c4000f533904a0b31df0f1",
+ "docs/report/latest-figures/paired-pilot.svg": "4b4dea6d5f8c9d11cfcc8611ab1e9a7beb147c07ea24a18e6adb12359c7aac54",
+ "fleet_witness/web/app.js": "69f16f6e1a04d91f2ff7343b1394d962795998889b72b25bab189042fa9cb1b0",
+ "fleet_witness/web/index.html": "a76ad03dc14fa2824b1a7b5223d59c0111e060de12a6bee6138ed9f1047865c3",
+ "fleet_witness/web/styles.css": "a81a1c81281d6f0de3ce90742107386ecc8e29c90c0b01f306d5ac0170382072"
+ },
+ "style": "Edited sequence of 16 browser states and caption cards; no audio, no continuous real-time capture.",
+ "task": "submission-finalization-20261109",
+ "tools": {
+ "chrome": "155.0.8059.39",
+ "ffmpeg": "9.0.2",
+ "node": "26.11.1"
+ },
+ "video": {
+ "format": {
+ "duration": "180.000000",
+ "size": "2506489"
+ },
+ "programs": [],
+ "stream_groups": [],
+ "streams": [
+ {
+ "codec_name": "h264",
+ "height": 900,
+ "pix_fmt": "yuv420p",
+ "r_frame_rate": "25/1",
+ "width": 1600
+ }
+ ]
+ }
+}
diff --git a/docs/superpowers/plans/2026-11-09-submission-finalization.md b/docs/superpowers/plans/2026-11-09-submission-finalization.md
new file mode 100644
index 0000000..28c768a
--- /dev/null
+++ b/docs/superpowers/plans/2026-11-09-submission-finalization.md
@@ -0,0 +1,126 @@
+# Submission finalization implementation plan
+
+> **For agentic workers:** Use superpowers:executing-plans. One fresh independent whole-branch review follows implementation.
+
+**Goal:** Make the submission easy to verify and demonstrate, then prepare its final event entry.
+
+**Architecture:** An offline standard-library verifier checks the explicit public
+package and derived data. Local browser capture produces labeled media without
+adding authority to the application. Existing CI and publication tooling integrate
+these additions.
+
+**Tech Stack:** Python 3.11+ standard library, existing Node runtime matrix, installed
+Chromium and FFmpeg for authoring only.
+
+**Spec:** `docs/superpowers/specs/2026-11-09-submission-finalization-design.md`.
+
+## Global constraints
+
+- Owner/integrator: Alejandro Castellanos. Task: `submission-finalization-20261109`.
+- One isolated worktree; retained artifacts in the admitted sibling task packet.
+- Preserve title/authors/abstract, numerical findings, datasets, accepted ADRs,
+ old REPORT artifacts and current PDF bytes. No paid calls or cloud mutation.
+- Private contact/account/form records never enter Git, media or the ZIP.
+- Exact source/runtime evidence, full Python suite, independent review and green
+ hosted CI precede integration. Existing owner green-merge authority persists.
+
+## Review focus
+
+1. Unsafe ZIP members or filesystem links must be rejected without extraction.
+2. Internally consistent but stale/tampered package data must not silently pass
+ the repository comparison or frozen-cohort checks.
+3. Duplicate pair presentations, wrong totals and misleading summary labels must
+ fail even if checksums are recomputed.
+4. Media must show actual application output, correctly labeled, with no private
+ desktop/browser information or implication of a new live-model run.
+5. Submission readiness and actual event confirmation must remain distinct.
+
+### Task 1: Offline verifier and CI
+
+Owned: `scripts/verify_submission.py`, `scripts/build_latest_submission.py`,
+`tests/test_submission_verifier.py`, `.github/workflows/ci.yml`.
+
+Interfaces: `verify_submission(package: Path, repository: Path | None = None) -> dict`
+checks ZIP bytes; `verify_directory(directory: Path) -> dict` checks an extracted
+allowlist. CLI defaults to the repo ZIP with source comparison; the standalone
+packaged script defaults to its own directory. Explicit `--package` verifies a
+ZIP; explicit `--directory` verifies extracted files. JSON output and nonzero
+exit on invalid input. No writes or network.
+
+- [x] Write behavioral tests using actual ZIP bytes and independent expected
+ counts: 160 trials, 80 pairs, raw 18 invalid approvals / zero invalid denials,
+ explicit zero invalid approvals / 40 invalid denials, 30 effects, 68 attempts,
+ 28 authorization denials, ten duplicate no-ops, $0.2605526 provider charge.
+- [x] Observe failure before implementation; add bounded readers, exact checksum
+ coverage, provenance/metadata consistency and cohort recomputation.
+- [x] Verify corrupt/missing/stale/duplicate/traversal/symlink/oversized input and
+ unchanged output trees. Add the verifier to the Python CI command sequence.
+
+### Task 2: Public media and package
+
+Owned: `fleet_witness/web/index.html`, `docs/assets/fleet-witness-cover.png`,
+`docs/assets/fleet-witness-viewer.png`, `docs/submission/latest/`,
+`docs/submission/latest-submission.zip`, `docs/report/latest.provenance.json`,
+`docs/submission/media/`, `README.md`, `docs/SUBMISSION.md`, `docs/AI_ASSISTANCE.md`.
+
+- [x] Correct the stale footer, generate the seed-42 demo and capture its real UI.
+- [x] Render and inspect the PNGs; record and review a captioned 180-second MP4,
+ including real UI interactions and separately labeled recorded study results.
+- [x] Record media source/digests/tools, update metadata and copy-ready links,
+ refresh publication provenance and rebuild the explicit package.
+- [x] Run the new verifier on the repo ZIP and extracted package, including its
+ standalone entrypoint. Inspect duration, frames, captions, file sizes and hashes.
+
+### Task 3: Acceptance, review and delivery
+
+Owned: `docs/BUILD_STATUS.md`, this plan and its spec; private artifact packet.
+
+- [x] Full Python/Node tests, planner, demo, links/whitespace and preservation checks.
+- [ ] One independent whole-branch review; owner resolves findings and runs the
+ exact final-source runtime matrix. Merge green work and synchronize clean main.
+- [ ] Prepare the form using user-provided data. Required missing information
+ stays pending; no invented contact details or submission success. Preserve the
+ actual confirmation if the form can be submitted.
+
+## Execution ledger and rulings
+
+- Native worktree tool has no path parameter; Git creates the worktree under
+ Projects as required by owner policy. Existing unrelated worktrees stay intact.
+- The brainstorming/design menu is unnecessary: the owner selected all three
+ concrete recommendations and previously requested autonomous execution.
+- Preserve the sibling artifact packet instead of deleting skill scratch.
+- Initial preflight used an unsupported flag and failed before heavy work;
+ corrected --budget-gib/--output-dir admission passed. Both records are retained.
+
+- Task 1 complete locally: eight red-to-green mutation tests; repository and
+ standalone package checks pass. The full suite now has 317 tests.
+- Task 2 complete: two inspected PNGs, 180-second H.264 video (2,506,489 bytes),
+ captions, public hashes and explicit 17-member package. PDF and data unchanged.
+- Task 3 local checks complete: 317 Python / five Node tests; planner, demo,
+ video decode and local links pass. Independent review and hosted CI follow.
+- Ruling: keep the existing admitted sibling packet as the execution ledger and
+ retain it after integration, instead of deleting skill scratch — owner policy
+ requires failed receipts and rollback evidence; this consumes bounded disk space.
+- Ruling: use an explicitly labeled edited sequence of captured browser states,
+ held for reading, with captions and no audio — it fulfills the walkthrough while
+ avoiding a real-time-execution claim; it is less fluid than a continuous recording.
+- No form submission: the owner is unsure of contact emails and prior submission
+ status. The private form record preserves that uncertainty without inventing data.
+
+- Independent review of 5ff5301 found three important verifier gaps: rehashed
+ frozen data/interpretation changes, NUL-suffixed ZIP names, and uncaught
+ decompressor errors. Three added regression methods failed first (six failed
+ subcases and one decompression error), then all eleven verifier tests passed.
+- Owner fix pass: validate original ZIP names, allow only stored/DEFLATE members,
+ sanitize zlib errors, check contrast labels and pin the accepted public table
+ after independent recomputation. New cohorts need a separately reviewed revision.
+- Review checklist finding regraded as important evidence accuracy: integration
+ remains unchecked until the final matrix and merge are observed. This is a
+ status correction, so no prose-matching test is added.
+
+- Reviewed-source acceptance: 320 Python / five Node tests passed; all four
+ independent malformed archives now yield sanitized JSON failure; package
+ repo/standalone checks, planner, links and full video decoding pass.
+- A 2 GiB authoring-manifest rerun was refused before testing. With media work
+ finished, a separate 512 MiB qualification-only estimate passed fresh admission;
+ both receipts retained, no reservations or thresholds changed.
diff --git a/docs/superpowers/specs/2026-11-09-submission-finalization-design.md b/docs/superpowers/specs/2026-11-09-submission-finalization-design.md
new file mode 100644
index 0000000..9bd356a
--- /dev/null
+++ b/docs/superpowers/specs/2026-11-09-submission-finalization-design.md
@@ -0,0 +1,39 @@
+# Final submission improvements
+
+Task `submission-finalization-20261109`; owner/integrator Alejandro Castellanos,
+assisted by Codex. Base: `2ce59d34f92516cfb6666b891a4b35da2b46c9ba`.
+The owner accepted the three proposed improvements and requested autonomous work.
+
+Add a standard-library, read-only submission verifier, a real viewer screenshot
+and PNG cover, and a three-minute captioned walkthrough using the existing script.
+Keep scientific results and experimental source identities frozen. Preserve prior
+packages, manifests, failures and rollback assets. No new inference or cloud work.
+
+## Design choices
+
+- A committed verifier replaces the copy/paste-only verification path. It accepts
+ the repository package or an extracted package, checks exact allowlisted members,
+ bounded regular-file/ZIP inputs, checksums, current provenance identities,
+ metadata/abstract agreement and independently recomputed pilot counts/costs.
+ It reports consistency, not independent provider authenticity. It never extracts
+ an archive, executes bundled code, contacts a network or repairs input files.
+- Check the published package in every Python CI lane. Meaningful mutation tests
+ must fail for stale source/package bytes, duplicate/unsafe ZIP members, oversized
+ or linked input, inconsistent provenance and incorrect pilot aggregates.
+- Render existing SVG artwork into a PNG cover. Capture the real local viewer
+ with a visible scripted label. Correct its stale solo-entrant footer before
+ capture; no application behavior changes.
+- Record a captioned 180-second browser walkthrough: mission, build-only evidence,
+ revocation/duplicates, recorded live-pilot results and reproducibility. Keep
+ synthetic replay and recorded model evidence labeled separately. No claim of
+ live cloud availability or new model execution. Use installed Chrome/FFmpeg;
+ publish MP4 separately to avoid inflating the portable evidence ZIP.
+- Include a standalone verifier and two PNGs in the explicit package allowlist.
+ Preserve the previous manuscript and existing PDF bytes. Update metadata and
+ links only after the media actually exists and has been reviewed.
+
+A new model study or viewer redesign would require broader review and would not
+help this final delivery. The accepted approach is additive presentation and
+verification using existing mechanisms. Contact details and form state stay in
+an owner-private artifact record, never Git. Submit only after required details
+are supplied and a concrete final package exists; retain the confirmation.
diff --git a/fleet_witness/web/index.html b/fleet_witness/web/index.html
index 41b645e..019430a 100644
--- a/fleet_witness/web/index.html
+++ b/fleet_witness/web/index.html
@@ -44,7 +44,7 @@
-
+