v0.3.0: verifier evidence and measured outcomes
还没有人认领这个 Issue。
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 20/100
- Issue 类型
- 功能
- 描述清晰度
- 基本清楚
- 活跃度
- 冷清
- 技术栈
- playwright, typescript
- 领域
- testing-qa
调研方向
The issue points to the full plan in the usabl-plans repository at 09-verifier-evidence-and-outcomes.md; start by reading that plan and the repository's existing gate and voicing-lane entry points. The scope spans six evidence and measurement mechanisms plus wiring the voicing lane, and work is held until v0.3.0 opens. Done means implementing the plan while preserving the stated verdict and receipt invariants.
由索引模型根据 Issue 内容生成。
描述
Tracking issue for the v0.3.0 verifier work. The full plan is held in the
usabl-plans repository (09-verifier-evidence-and-outcomes.md); this issue records
the scope so it is not lost. No work starts until v0.3.0 is opened.
Scope
v0.3.0 extends usabl from a gate that proves the present into one that also
publishes evidence about its own sensitivity and measured outcomes. Six
mechanisms, plus wiring the voicing lane:
- Task-start proof brief: a machine-readable brief an agent reads before a task,
declaring what must be proven and what counts as done. - Sensitivity evidence: a mutation catalog and runner that inject declared
defects and show the gate detects them, with an attestation and two
sensitivity levels. - Published Orca calibration: compare predicted announcements against Orca on a
fixed corpus, with a calibration attestation, a promotion rule, and staleness
handling. - Observed barrier-days: a ledger measuring how long real barriers persist, as
measurement only, with no verdict and no receipt. - Evidence-bound assistive-technology docs: publish AT behavior docs bound to
the evidence that produced them. - Agent self-correction study: measure whether agents repair barriers when the
gate tells them what is wrong.
Plus: wire the voicing lane, which is built, exported, and currently unwired in
v0.2.0.
Non-negotiables carried from the engine
The gate stays the only verdict authority. Measurement surfaces mint no verdict
and no receipt. Honesty invariants do not flex: a wrong guess is a gap or a
refusal, not a verdict.
Status
Held for v0.3.0.
- 主要语言
- TypeScript
- 星标
- 0
- 派生
- 0
- 平均合并
- 2 小时 48 分钟
- 30 天内合并 PR
- 8
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 没有贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
usabl-dev/usabl 的其他 Issue
-
area: engine bug
难度 2/5 1-3 小时 新手友好度 86/100
-
Stop hook allows a stop on an unknown verdict, and the self-check prints IDLE for a null verdict未关闭area: assistants bug
难度 2/5 1-3 小时 新手友好度 75/100
-
area: engine bug security
难度 2/5 1-3 小时 新手友好度 72/100
-
Overlay: browser context failures are counted per plugin, not per browser, and close is not shared未关闭area: overlay bug
难度 2/5 1-3 小时 新手友好度 64/100
-
area: engine enhancement performance
难度 2/5 1-3 小时 新手友好度 68/100
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 76/100
维护者通常 1 天内回复
-
bug
难度 1/5 1 小时以内 新手友好度 76/100
维护者通常 1 天内回复
-
community first-timers-only good first issue hacktoberfest help wanted low hanging fruit up-for-grabs
难度 1/5 1 小时以内 新手友好度 76/100
lingdojo/kana-dojo#32964 · 1 条评论 · 5 个 reaction ·
维护者通常 1 天内回复
-
Documentation status: needs triage
难度 2/5 1-3 小时 新手友好度 72/100
mastra-ai/mastra#28194 · 1 条评论 ·
维护者通常 1 天内回复
-
priority-low refactor-3
难度 2/5 1-3 小时 新手友好度 84/100
维护者通常 3 天内回复