Reported issues for arkgate
Pod holds 6 of 6 GitHub reports that passed its relevance review. This can include external user reports, maintainer-confirmed bugs, and concrete feature gaps. Treat them as evidence to inspect, not a count of distinct defects.
Back to arkgate.
Most discussed
Dogfood: check exits 0 while printing ✖ type-only LAYER_IMPORT_VIOLATION
Summary
Field dogfood against an anonymous third-party TypeScript monorepo with published arkgate@4.8.13.
arkgate-check printed a red ✖ LAYER_IMPORT_VIOLATION for type-only placement debt, then reported overall pass with exit 0 ("Ark check passed with 1 type-only placement debt (non-blocking…)").
Why it matters
Pass/fail signal is split: the glyph looks like failure; the exit code (and top-line “passed”) says success. Anyone skimming output will misread the result; anyone…
Read the thread · 2026-09-09 · closed · 2 comments
Dogfood: start vs --recommend disagree on projected governed coverage
From live dogfood (anonymous third-party tree)
Same clone, same moment, same crud-product / hexagonal shape signals:
| Surface | Claimed projected governed coverage |
|---|---|
arkgate start (preview + refuse stderr) |
36% |
arkgate-check --doctor after apply |
36% (34/94) |
arkgate-check --recommend |
89% (below 90% confirmation warning) |
Both start and recommend label the number as projected governed coverage. A ~53pt…
Read the thread · 2026-09-12 · closed · 1 comment
Dogfood: start --apply refuse gate missing from one-minute path
From live dogfood (anonymous third-party tree)
README Start in one minute lists:
npx arkgate start
npx arkgate start --apply
On a medium brownfield TypeScript product tree (UI + data), bare arkgate start --apply exits 2 with:
- projected governed coverage below 50% → refuse apply
- archetype confidence below 0.6 with coverage < 80% → refuse without explicit shape
Refuse-closed is correct product behavior. The bug is first-contact docs / UX: a smart stranger…
Read the thread · 2026-09-12 · closed · 1 comment
Write hook denies imports that ark-check allows: layer resolution ignores explicit file patterns when a broader glob also matches
Written by a Claude Code agent run by @pedroknigge, from a real session on a private repo (Amarilla Platform, arkgate
4.8.14). Everything below was measured in that session, not inferred.
Summary
The write hook (arkgate-mcp --hook) and ark-check disagree on layer resolution when a path is matched by both an explicit file pattern and a broader glob. The hook resolves by the broad glob; ark-check gives precedence to the explicit file pattern. So the hook denies writes…
Read the thread · 2026-09-08 · closed · 1 comment
P2: library monorepo start emits Next.js-flavored layer copy
Field dogfood (2026-09-05 ART)
Target: colinhacks/zod (TypeScript-first schema library monorepo, not a Next app). Command: arkgate start / start --apply (arkgate@4.8.11).
Observed
- Shape detection: "Several deployable units in one repository." (reasonable for packages/*).
- Generated ApplicationOrchestration / PresentationAdapters descriptions talk about Next App Router API, Pages API, app/api, etc.
- Doctor coverage lists those Next-flavored blurbs even though the tree has no Next…
Read the thread · 2026-09-06 · closed · 0 comments
P2: doctor cold path ~2.4m on medium TS monorepo (zod)
Field dogfood (2026-09-05 ART)
Target: colinhacks/zod (~27MB git size, 226 governed TS files after start). Command: arkgate-check --doctor and --doctor --json with typescript-ark-host available.
Observed
- Cold doctor ~142s wall clock (twice: human doctor + JSON).
- Without TS host, doctor fails fast with Analysis unavailable (exit 2) — good fail, but stranger path depends on install succeeding first.
Why it matters (Guiar)
Doctor is the control plane. Multi-minute first status…
Read the thread · 2026-09-06 · closed · 0 comments
Most recent
The remaining reports are on the project's issue tracker.