Repository navigation
STE-58: the demo video and the flow screenshots, recorded against production - #70
Merged
Merged
Conversation
The two evidence items the SOW asks for under Deliverable 3, and the only ones the Instawards report still listed as "not yet produced". The video is 2 min 54 s of the real loop against production: a skill registered and audited on camera, its verdict and VERIFIED badge written on chain, an agent paying 0.1 USDC over x402 and receiving a soulbound licence, a borrowed licence refused with 401 until the wallet signs, and a poisoned skill landing DANGEROUS and being refused at purchase by the contract itself. Nothing is re-enacted. The seven transactions shown were checked against Horizon after the take: 7 of 7 successful, ledgers 4916394 to 4916409. Three things were added in the edit and they are named in the README: the typing animation, the caption strip, and the cards. Idle gaps were trimmed to 1.2s, so the audit that takes 22 seconds appears in ten — the raw .cast files carry a timestamp per event, so anyone can check the real pace. The numbers on the closing card were re-measured the same day rather than copied forward: 13 of 13 catalogue skills SAFE, 4 of 4 poisoned samples caught, 36 of 36 report hashes matching, 15 settlements each with a licence minted. The README maps every chapter to the transaction or endpoint it shows, lists what each screenshot proves, and states plainly what the video does not claim: submission is still operator-run, every wallet belongs to the team, and the model never decides a verdict. The file is committed once and served twice — docs/evidence/demo/sterish-demo.mp4 for review, landing/public/demo.mp4 so it plays at sterish.xyz/demo.mp4. Git stores identical content as one blob, so the second path costs nothing.
… anchors `verify_onchain.py` reads 36 reports and matches all 36 against their on-chain evidence hash, but only 25 of them were in the repository: the other eleven came from demo and rehearsal runs and had never been committed. That gap matters, because a reviewer who clones the repo and runs the command gets 25, not the 36 the report claims. These are the reports behind the demo runs, including the two skills recorded for the video. Nothing is regenerated — the bytes are exactly what the pipeline wrote and what the evidence hash on chain commits to, so the claim is now reproducible from a clean clone.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The last two SOW evidence items. The Instawards report has been carrying two bold "Not yet produced" rows since it was written; this removes both.
What is in here
docs/evidence/demo/sterish-demo.mp4— 2 min 54 s, 1080p, no audio. Also atlanding/public/demo.mp4so it plays at https://sterish.xyz/demo.mp4.docs/evidence/demo/screenshots/— nine full-page 2× captures of the live product.docs/evidence/demo/*.cast— the raw asciinema recordings, replayable.docs/evidence/demo/README.md— the evidence map.Recorded, not re-enacted
The take registered a skill, audited it, bought it and then poisoned a second one, live against
api.sterish.xyzand testnet. The seven transactions on screen were re-checked on Horizon afterwards — 7 of 7 successful, ledgers 4916394 → 4916409:Three things were added in the edit, and the README names all three: the typing animation of each command, the caption strip, and the cards. Idle gaps were trimmed to 1.2 s, so a 22-second audit appears in ten — the
.castfiles carry a timestamp per event, so the real pace is checkable rather than asserted.The numbers on the closing card were re-measured, not carried forward
GET /skills?include_test=truecom.fixtures.poisoned.*verify_onchain.pytotal_supplyon the Tokens contractWhy eleven report files came with it
verify_onchain.pyread 36 reports and matched all 36, but only 25 were committed — the rest came from demo and rehearsal runs. A reviewer cloning the repo and running the command got 25, not the 36 the report claimed. Now the claim reproduces from a clean clone.What the video deliberately does not claim
Submission is still operator-run (STE-56), every wallet in it belongs to the team, the model never decides a verdict, and it is testnet only. All four are stated in the README rather than left for someone to notice.
Note for @Nabil-Aufa and @AncungAulia
One file lands in your app:
landing/public/demo.mp4, a static asset, nothing else inlanding/touched. Committed once and referenced from two paths — git stores identical content as a single blob, so the second path costs no space. This complements STE-28, which stays yours for the polished launch cut on X; this is the evidence cut for the grant.Linear: STE-58 · cc @axel @fable