444060c3eb
- Introduced `critic`, an independent adversarial reviewer for security and correctness. - Added `fable-orchestrator` to manage task routing and verification. - Implemented `gauntlet-critic` for fresh-context evaluation of gauntlet rounds. - Created `planner` for generating executable implementation plans with dependencies. - Developed `security-auditor` for application security reviews and audits. - Established `system-steward` to improve agent prompts and skills based on verified failures. - Added `dev-loop` skill for autonomous development loops over repositories. - Implemented `gauntlet-loop` skill for iterative quality benchmarking against reference standards. - Updated project settings to utilize the new orchestrator agent. - Created documentation for `GAUNTLET.md`, `PROGRESS.md`, and `REFERENCE_BAR.md` to track project status and quality benchmarks. - Added detailed prompting style guide to enhance understanding of prompt patterns and agentic loops.
3.3 KiB
Reference bar
The concrete quality bar for gauntlet work. Every entry must point at something a referee can open, run, or look at — an adjective is not a bar. Changing a bar mid-gauntlet is an owner decision recorded in
DECISIONS.md.Seeded 2026-08-06 at the gauntlet-loop/fable upgrade. This project already has a real bar: the interactive prototype + the Nocturne token authority + per-screen contracts. Precedence guard (D-2026-07-31-01 lineage): the prototype is evidence, never authority — where the prototype and the recorded spec disagree,
08-development-spec > 04-rules > PRDwins and the difference is not a gap. The referee grades against the spec-corrected prototype.
Base references: PROTO = PS Bus Ticketing App - Conductor App.html (repo root — open in a browser, navigate to the screen) · TOKENS = docs/06-ui-patterns.md (Nocturne) · SPEC = docs/08-development-spec.md (per-screen contract) · DESIGN = docs/design/** (screen specs, where written).
Bars by part
One row per screen/flow as it enters a gauntlet — seeded with the screens that already have design artifacts; add rows using the template as work reaches each screen. Budgets live on the GAUNTLET.md board.
| Part | Reference artifact(s) | How to compare | Minimum parity |
|---|---|---|---|
| Auth screens 1–2 | PROTO auth screens · docs/design/ auth spec · SPEC §screen criteria |
run the app on the 2 GB reference device (or emulator at its profile), screenshot vs PROTO side by side; check tokens vs TOKENS | layout/hierarchy/tokens match the spec-corrected prototype; per-screen SPEC criteria pass |
| Screen 06 — discount capture (dual-photo) | PROTO screen 06 · docs/design/ screen-06 spec · SPEC criteria |
walk the capture flow on-device; screenshot each state | every state (capture, retake, proof review) present and one-handed operable; ≥ 48 dp targets |
| Screen 11 — printer setup | PROTO screen 11 · docs/design/ screen-11 spec |
walk pairing/test-print flow (or its no-hardware stub — see orchestrator memory: no printer hardware) | states + error paths match; no-hardware path explicit, never silent |
| P10 — prepaid booking / QR | PROTO P10 · docs/design/ P10 spec · SPEC criteria |
walk the flow offline; screenshot | offline-first behavior + states match the spec-corrected prototype |
| [next screen] | PROTO screen NN · docs/design/ spec if present · SPEC criteria |
on-device screenshot side-by-side + flow walk | [what must match] |
Behavioral bars (not screenshots): the ≤ 20 s record-a-passenger contract (stopwatch on the reference device), 7-day-offline invariants (A-1…A-6), and the TC-* tables in docs/09-test-plan.md — these are already acceptance tests; the gauntlet adds the visual/UX parity layer on top, it does not replace them.
Reference sources
PS Bus Ticketing App - Conductor App.html— interactive prototype (root)docs/06-ui-patterns.md— Nocturne tokens/components (authority for visual language)docs/design/**— written screen specs (authority over the prototype)docs/08-development-spec.md— per-screen acceptance criteria
Out of scope for the bar
- Anything the recorded spec has changed from the prototype (spec wins; log the delta as evidence, not a gap).
- Server/back-office UI (contract-only,
docs/07-api-contract.md), iOS, passenger-facing surfaces.