Home Graphsmith
Graphsmith
A design-time environment for graph-based AI and automation workflows.
Scorecard
82/ 100
Stage: Beta on our own rubric. Google Play: Web only.
Primary journey: shown with sample data
Scored 2026-11-08 on our own rubric; repository at commit b9aca48.
What we promised, and what we have not finished. We promised one public rubric for every product, a list of what is not finished on every page, and a named source for every number. Most of these products are not finished, and each page says where it stops. Read the rubric or every claim and its source.
Recorded 2026-11-08. The scorecard in the video shows 67; the site shows 82/100 on our own rubric (2026-11-08, per-feature rule D32). The phone-size fix (#28) is not shown. Re-recorded 2026-11-08 against graphsmith main b9aca48; adds the State tab tagging a reviewer decision chosen in Evaluate as user, not mocked (#66), shows the readiness coverage tiles (#75), and the score, how-checked and gaps cards are reworded (score 67 as published, newer audit pending). The phone-size tab row fix (#28) is not shown. Voice is AI-generated (ElevenLabs). Captions timed by forced alignment, 24 cues, worst cue 0.1 s from speech; frame scan found no keys, addresses or paths. Voice is AI-generated (ElevenLabs). Shows made-up demo data. Read the transcript.
This button opens your email app with the subject already filled in. Your message comes to support@shipitworks.com. There is no form and no database. Privacy notice.
Works today
Each line links to the scorecard row that backs it.
Not finished
From the same scorecard, next to the claims.
- Headless Chrome checks (check:ui 14/14, check:display 42/42) ran on one machine on 2026-11-07 and 2026-11-08; the read-only share view and other browsers are not covered, and no screen reader has been run
- graphsmith.dev was fetched once on 2026-11-07 and served the build checked then; a 2026-11-08 check:display run against it failed 4 of 42 (phone tab row), fixed on main in PR #28. Whether the live site serves b9aca48 is not yet traced; uptime was not measured
- A request capture of one journey on graphsmith.dev (2026-11-07) saw 0 requests after load and 3 during load; it does not cover browser extensions or other browsers
- npm audit ran from an offline cache, so the 0-vulnerability result is not independently confirmed
- A privacy page exists (#/privacy) and its network, storage and redaction sentences are pinned to the app code by tests; whether it meets a legal privacy-policy requirement was not checked
- CI on main head b9aca48 was queued, not green, when checked on 2026-11-08; last green run is 8b83b34
- package.json says 0.1.0 while the release tag is v0.4.0
- The independent proof pack (proof/proof-pack.json) is from 7c913f0 on 2026-11-02 with verdict STOP; its share-link defect is fixed by PR #20, but the pack was not regenerated, so test counts (153) are from that audit, not from b9aca48
- Proof logs cite the commit they ran on; except the three feature logs (bfa6b4d) those commits are branch tips replaced by squash merges, so they cannot be opened in main's history
What changed
Dated lines from the scorecard. The source code is private, so these lines are the public record.
- 2026-11-08: Release v0.4.0 on main head b9aca48. CI moved to the self-hosted runner (PR #107). Recorded browser run of the readiness report, coverage, State map, decision rules and connections: 28/28 on three templates, the control fails (PR #75, proof/readiness-state.log). Lane notes added to FOUND.md (PR #31).
- 2026-11-08: Recorded runs for the three scorecard features: domain logic 107 tests, workspace UI 14/14, live demo build match (PR #8); network capture on graphsmith.dev, 0 requests after load, typed text in no request (PR #7). The three features are now Proven.
- 2026-11-08: Phone-size workspace tab row now wraps; a new check runs every route at phone size and in dark mode, 42/42 (PR #28).
- 2026-11-08: The State tab tags a reviewer decision chosen in Evaluate as user, not mocked, and the second-visit auto-approval as generated; the tab's note explains all four tags (PR #66).
- 2026-11-08: Share links cap rules at 20 per decision and 400 in all, and clip quoted decision text to 80 characters (PR #20). Readiness checks carry the NIST AI RMF 1.0 subcategory each relates to; the mapping is a judgement, not measured (PR #21).
- 2026-11-08: Privacy page and in-app copy pinned to the code: network and storage sentences (PR #10), the three page-load requests (PR #12), exclude fields removed not replaced (PR #27, PR #29); landing page templates and checks lines (PR #9, PR #26).
- 2026-11-07: The readiness checks are tagged with the NIST AI RMF 1.0 subcategory each relates to (the subcategory text is quoted from NIST; which subcategory a check relates to is a judgement, not measured). Share links now cap rules and clip quoted decision text, fixing a defect found in the proof pack, and the export label, share toast and Privacy page say that exclude fields are removed from scenarios, not replaced. The Privacy page lists the three requests a browser makes at load.
- 2026-11-07: A new check runs every route at phone size and in dark mode, and fixed a phone tab-row overflow. Recorded runs on one machine in headless Chrome: canvas, inspector, outline and tabs; no off-site request in the demo journey; the live demo served the build checked then; unit tests for simulator, lint, export, share and drafter. Release v0.3.0.
The next step for Graphsmith.
Email us about Graphsmith