You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Make an ordinary Agent's path from a bounded system description to a readable, verified diagram more reliable and less expensive to repair. Use the existing issues below as the implementation owners, with one selected approach per behavior.
Assessment baseline: 920543baa1c6137803c5b45a69d8977152773d35, checked on 2026-09-08. Local reproduction used that commit's tracked archify.zip (2.17.0-dev.1) on macOS, Node v26.3.0, and Chrome. The CI run for the same commit passed.
Current maintenance status — 2026-09-08
This section records the current maintainer decisions. The assessment below remains historical evidence from its stated baseline, not a description of current main.
#256 remains open at 502c22e; 12 CI checks and CodeRabbit status pass. Independent collaborator approval is still required.
Do not bypass last-push approval. Triage the two latest bot suggestions about package-wide font detection and positive staging coverage before final acceptance; a green bot status is not a resolved review thread.
Deferred by maintainer decision; issues remain open.
Do not change default font sizing or add a typography-scale contract in this workstream. Retain the existing Viewer zoom controls. This is a scope decision, not a claim that zoom proves default readability is satisfactory.
#175 / #214 / #336 generation efficiency and repair rounds
Open; not established as solved. #353 has collaborator approval and passing CI at 62e7c4d.
#353 is a bounded workflow-overflow diagnostic improvement. Its Luna trial failed before the new guidance was exercised, so it does not demonstrate fewer repair rounds. Rebase/revalidate against current main before integration, then evaluate matched tasks while retaining failures and timeouts. Do not close these issues from diagnostic tests alone.
#344 stays open. Implementation merged into main is distinct from a published release; a deferred item is distinct from a completed fix.
Evidence and proposed sequence
Priority
Work item
Existing owner and acceptance direction
P0
Credential-safe repository evidence
#332 — same-repository credentialed HTTPS must verify, and every diagnostic surface must redact credentials. Independent reproduction.
P1
A usable installed starting point
#324 — all five bundled examples must pass the required real-browser viewport checks. Independent reproduction. Preserve meaningful nodes, relationships, labels, cards, and the existing quality gates.
P1
Earlier feedback and fewer repair rounds
#175, #214, #336 — expose usable layout/viewport budgets earlier and provide evidence-based repair guidance; compare matched tasks before selecting a solution.
P1
Comfortable default typography
#340, #282 — evaluate font size, node dimensions, wrapping, and canvas layout together, with real Chinese and English content and browser review.
P1
Portable evidence and offline delivery
#301/#143 and #242/#277 — separate local Git verification from provider-specific source links; select and test a predictable offline font strategy.
P2
Bounded Viewer maintenance
#343 — extract one cohesive source capability while retaining deterministic, single-file delivery and installed-Skill compatibility.
Specific observations behind this ordering:
Three packaged examples fail all four desktop containment checks. Fresh Data Flow delivery still reports 9/9 with zero errors and warnings and produces the exact same failing HTML. Deterministic delivery, browser evidence, and perceptual review need distinct results throughout the workflow.
The current projected node-text floor is 6px. In the audited default Architecture example, the minimum measured node text was 8.65px. Passing the existing floor does not establish comfortable reading.
The extracted ZIP's five examples pass real Chrome checks at 1440 × 900, 1600 × 1000, 1920 × 1080, and 2048 × 1320, retaining light/dark captures and separate perceptual review.
Fixed tasks, model settings, source revisions, and package identity are recorded before the convergence comparison. Keep failed and timed-out runs in the results.
Report first-pass usability, repair-round distribution, and time to a verified artifact separately; demonstrate improvement without losing required semantic content or visual quality.
Objective
Make an ordinary Agent's path from a bounded system description to a readable, verified diagram more reliable and less expensive to repair. Use the existing issues below as the implementation owners, with one selected approach per behavior.
Assessment baseline:
920543baa1c6137803c5b45a69d8977152773d35, checked on 2026-09-08. Local reproduction used that commit's trackedarchify.zip(2.17.0-dev.1) on macOS, Nodev26.3.0, and Chrome. The CI run for the same commit passed.Current maintenance status — 2026-09-08
This section records the current maintainer decisions. The assessment below remains historical evidence from its stated baseline, not a description of current main.
b86b607); both issues closed.502c22e; 12 CI checks and CodeRabbit status pass. Independent collaborator approval is still required.62e7c4d.Active sequence
#344 stays open. Implementation merged into main is distinct from a published release; a deferred item is distinct from a completed fix.
Evidence and proposed sequence
Specific observations behind this ordering:
Scope and contribution coordination
Acceptance evidence