verifymonochrome · nothing moves

Boomship/Verify

Done is a claim.

Prove it.

Mission 0218Illustration

Launch Northstar v4

Research

sources reviewed31

internal documents4

contradictions surfaced2

Product

files changed4

tests27 / 27

browser flows5

lintnot run

Artifacts

launch pageverified

board deck12 slides · PDF

research memo4 claims · cited

campaign brief1 page

Operations

scheduleFriday · 5:00 PM

steer receivedturn 2

decisionrecorded

Security

reviewno high-risk findings

secrets0 in added lines

Independent review

reviewera model the builder did not use

verdictpassed

Model

cost$1.86

time48m 12s

VerifiedVerified where possible · every check named · not run stated

A receipt like this ends every Mission. Every row is a check that ran, or says it did not.

act 01 · the verifierdid not do the work

Act 01The verifier

A verifier that did not do the work.

After the builder stops, a separate stage observes the working tree on its own: the diff, a fresh test run, a typecheck, a review by a different model. It does not read the builder’s summary. Each line below is a step on the ledger.

verifier logmission 0218complete · 01:21

00:00.00git diff HEAD · 4 files · +61 −19

00:00.08semantic diff · PricingHero() changed · HERO_COPY added

00:00.41secret scan · added lines · 0 findings

00:01.02predicted tests · 2 files from the import graph · 6 / 6

00:03.18npm test -- --silent · 27 / 27 · 1.9s

00:09.77tsc --noEmit · clean

00:09.78lint · not run · no lint script in package.json

00:31.40independent review · model ≠ builder · pass · 0 findings

00:52.07security review · high risk · pass · nothing exploitable

01:04.90visual regression · desktop, mobile · 71 → 86 · 0 new issues

01:21.33browser · 5 flows · nothing broke

01:21.35receipt written · confidence high

every line is a step · hashes of input and output on eachIllustration

act 02 · the checksnot run is written down

Act 02The checks

Every check named. Not run is written down.

Eight checks. A check that could not run is on the receipt as not run, with the reason, never left out. Before a Deep or Swarm run an acceptance contract sets the risk; high and critical risk require the security review.

testsThe project’s tests, re-run by the verifier. Tests the import graph predicts for the change run first.passed27 / 27

typechecktsc --noEmit in the workspace.passedclean

lintThe project’s linter, when it has one.not runno lint script

reviewAn independent model reviews the diff. Never the model that built it when another is eligible.passed0 findings

securityAn adversarial review of the diff. Required at high and critical risk.passedhigh risk

secretsA local scan of the added lines for credentials. git_commit refuses staged secrets.passed0 in added lines

critiqueA visual critique of the preview, before and after, when UI files changed.passed71 → 86

browserA real Chromium pass over the preview.passed5 flows

7 passed · 0 failed · 1 not runIllustration

act 03 · claims and evidencethe ledger · observed vs asserted

Act 03Claims and evidence

Claims on the left. Evidence on the right.

A step is observed when a real command, file or model result produced it, asserted when only the model said so. The semantic diff reads the change as symbols added, changed and removed, from both versions of every file.

“I updated the pricing hero in PricingHero.tsx.”

write_file · components/PricingHero.tsx5f20cc71observed
semantic diff · PricingHero() changed · HERO_COPY added18a6f3d9observed

“All tests pass.”

run_tests · npm test -- --silent · 27 / 27 · exit 0c41e88d2observed

“The change is safe.”

summary · the model’s own words44d19ab3asserted
security review · high risk · pass9d7e01f4observed

sha256 of input and output, first 16 hex, on every tool stepIllustration

act 04 · confidencerated from what ran

Act 04Confidence

Confidence is rated from what ran.

Four levels, each with its rule. The receipt states the level and the reason. The run is not done until the checks its risk requires have run.

HighTests passed on a fresh run, typecheck is clean, and an independent reviewer found no defects.

MediumTests passed, but the reviewer raised concerns or there was no review.

LowA required check failed, or a check the risk level requires did not run.

Not verifiedNothing could be observed: no runner, a failed or cancelled run, or no change to verify.

act 05 · prove it · undoa fresh receipt · the checkpoint before

Act 05Prove it. Undo.

Don’t trust the receipt. Re-run it.

Prove it runs verification again against the workspace and writes a fresh receipt. Undo restores the checkpoint taken before the run. Both act on the workspace, never on a summary.

mission 0218 · receipt 1verified high · 48m ago

tests27 / 27 · fresh run
typecheckclean
independent reviewpassed · model ≠ builder
lintnot run · no lint script

both act on the workspace, not on the modelIllustration

act 06 · boomship verify/verify <PR> [--report]

Act 06Pull requests

Boomship Verify for pull requests.

The same verifier, on a change nobody at Boomship wrote. Dependencies install with scripts disabled; tests run in a sandbox with no network and no token.

Read the docs

composer/verify https://github.com/northstar/web/pull/412 --report

fetchread with your GitHub token into a host-private mirror; the sandbox has no network and no token
resetsoft-reset to the merge base, so git diff HEAD is exactly the pull request
installnpm ci, lifecycle scripts disabled, every .npmrc removed
verifythe same verifier as a Build run, sandboxed
reporta Work Receipt; with --report, a “Boomship Verify” commit status

Boomship VerifyVerified high: tests, typecheck, review passedIllustration

verify somethingno field · no prism

What needs proving?

Start a Mission Security