Skip to content

task: Dashboard 补齐知识检索路径、排序解释与候选证据关联 #406

Closed

Description

@amy-why-3459

Current state: Proposed。当前 main 已有知识上下文详情、按轮次展示检索的 Run review,以及逐轮搜索/采用记录的导入能力;仍需补齐实际入口与排序过程的解释、共享记录的展示接入,以及知识到候选、改动和验证结果的完整导航。本 issue 是 #344 的具体展示与接入补充,关联 #379、#345、#392 和 PR #405。
Decision needed: 请维护者确认与 #344 的共享记录契约分工、首版交付切片及负责人;已合并的记录能力不等于 #344 或本 issue 已完成验收。
Next action: 复用现有 capture、optimization log、#384 的逐轮搜索记录及主线 Run review,核定实际依赖版本,提出向后兼容的数据补充与交互方案,再按验收项交付实现。

Goal and acceptance

让读者从看板回答两个问题:

  1. 这次知识搜索怎么发生的? 调用了哪个入口,从哪里搜索,如何匹配和排序,哪些内容因过滤、条数或上下文预算被排除?
  2. 搜索结果如何关联本轮优化? 哪些知识进入准备的上下文,哪些被声明采用,关联哪个候选/改动,以及后续验证结果是什么?

目标导航如下;连线只表达已记录的关联,不宣称知识造成了性能收益:

flowchart LR
    A["检索入口与请求"] --> B["来源、匹配与排序状态"]
    B --> C["命中、选中与上下文装配"]
    C --> D["声明采用与依据"]
    D --> E["轮次、候选与代码改动"]
    E --> F["质量、性能与证据"]
Loading

Current behavior and motivating gap

以下按 2026-11-13 核对的 main fe16b29388de29708d30ea388b589355540f3172 描述;实现前应再次核定依赖版本。

  • knowledge search 与 collect --relevance 是独立入口,共享打包知识记录的加载能力,不会自动串行调用。catalog.search_knowledge 要求查询词匹配,按出现次数排序,同分按条目 ID 排序。
  • collect 的相关性排序不能一概写成 skill 排序:weighted-lexical-v1 / weighted-raw-string-v1 面向 package skills,task-lexical-v1 面向 package skills 和 records。显式 IDs 路径保留调用者顺序并跳过相关性排序;未请求 relevance 时不执行相关性排序。展示应读取当次 selection.method/scope/ranking_applied 及请求记录,区分未请求排序、显式 IDs 跳过排序和旧记录未知。
  • campaign capture-context 通过 collect() 保存快照;传入 relevance 且未指定 IDs、也未覆盖方法时,默认选择 task relevance。已有快照保存 request.relevance_method、selection、items、omissions、context_omissions、字节预算及装配正文/哈希。当前 collect 内部计算排序分数,但返回条目没有暴露逐项分数或分解,capture 也没有保存这些分数。已有字段展示与新增执行点采集应分开交付。
  • 当前“优化过程复盘 → 知识上下文”已有请求、选中条目、遗漏原因、资源及哈希、拼接正文和 used_refs。主线 Run review 已能按 iterations[].context_ids 展示本轮检索,并显示已有 method/scope、返回顺序、可得匹配数及显式 ID 顺序;这部分经 Compare stage timings and inspect knowledge searches for each profile #401 合并,不再只属于 PR feat: Dashboard 实验性能曲线、事件轨道与证据下钻 #405 分支。无需重新实现已有详情。
  • feat: record per-round knowledge searches and adoption evidence #384 已合并,主线已有 campaign import-search、本地 search/round-<n>.json(schema v1)和可选的可移植 search/archive/round-<n>.json(schema v2)记录。它们提供命中、读取、声明/采用依据、覆盖状态及候选/run 关联,尚不能据此声称 task: 支持每个优化迭代的知识库命中归因统计 #344 全部展示与统计需求已验收。
  • 动机场景为 [Campaign] MiniMax-H3-Image on 4x910B1: match author workloads at >=50% H200 performance #390 / PR docs: record MiniMax-H3-Image Ascend iteration evidence #394 的 H3 优化。创建本 issue 时的本地只读观察发现:初始检索选中 5 项,预算内装配 2 组资源,其余 3 项因上下文预算不足被省略;used_refs 和结构化 iterations 为空,看板也未读入候选记录。因此当时已有实验的检索与采用关系无法逐轮导航。这是创建时的本地接入观察,本次未重新检查该私有历史;不应视作 PR docs: record MiniMax-H3-Image Ascend iteration evidence #394 已公开全部检索快照或可以重建未记录的历史。

Scope A — 检索入口与排序解释

  • 已有字段的只读投影: 在现有知识详情中补齐可展开的检索路径,复用 request、selection、items、omissions、context_omissions、预算和资源哈希。展示实际调用入口、包装调用与底层方法、来源类别及版本、查询/相关性提示、过滤范围、条数与上下文预算。明确区分 knowledge search、collect --relevance、campaign capture-context → collect,并按当次 selection 区分 skill-only、task-ranking、显式 IDs 跳过排序及未请求排序。缺少足够记录时显示未知,不用方法默认值猜测历史路径。
  • 缺失信息的执行点采集: 为新记录补充实际排序方法/版本、逐项可得分数与组成、并列排序规则,以及可记录的匹配、过滤和截断原因;返回顺序沿用执行结果。catalog 和 collect 已计算但未返回的分数应在各自执行点记录,并以向后兼容方式供 capture/展示消费。未采集到的候选或过滤详情仍显示未知,不要求重建完整历史候选集合。
  • 分开呈现“匹配/命中”“返回选中”“正文已装配”“声明采用”。没有请求正文装配时明确表示该阶段未执行,不能等同于预算淘汰。分数和解释只来自当次记录,不用当前算法、权重或知识正文倒算旧快照。
  • 复用并适当扩展现有检索返回与 capture 契约;记录可由调用者显式保存。保持原检索语义与排序结果,不要求每次执行两种入口,也不引入自动全局监听。新增观测应对照改动前后的返回引用、顺序及上下文装配结果,证明只增加解释字段。
  • 提供列表/表格与详情下钻,定位到知识条目、版本、资源路径和可用来源引用;不暴露未审核的私有原文或本地绝对路径。

Scope B — 知识到候选、改动和验证结果的关联

  • 复用 knowledge[].context_id/sha256/used_refs、iterations[].context_ids/knowledge_refs/candidates 及现有 candidate/run identity;并接入 feat: record per-round knowledge searches and adoption evidence #384 已交付的逐轮搜索记录。与 task: 支持每个优化迭代的知识库命中归因统计 #344 共用同一份命中、读取、声明采用、依据和结果关联,不在 Dashboard 另建采用判定或平行统计体系。
  • 按记录版本保留证据语义: 本地 v1 和可移植 v2 的采用口径不能混用。v2 的 declarations 保存调用者声明,materials[ref].evidence 保存经 review 建立的片段到决定的关联;显示依据类型、来源与 coverage。不得把旧版 used 或调用者声明自动升级为 v2 的已核验关联,也不能把 counts: null 当成零。使用现有解析和校验边界;版本、身份、来源或哈希不匹配时明确降级/提示。
  • 在 campaign 知识详情、Run review 和候选详情间提供双向导航:检索 → 条目 → 本轮声明使用/明确舍弃/未知 → 候选及可用 commit/diff/patch → 质量/性能结果与证据。复用 task: Dashboard 增强:每轮改动说明、动态 workflow 与耗时、测试进度 #345 的 campaign/iteration/candidate/run 身份,为同一候选的多次运行保留明确身份,不按显示名称或时间相近补关系。
  • 调用者声明采用与可核验的引用/操作/产物依据分开显示;上下文准备完成不证明模型收到或实际使用了内容。空 used_refs 仅表示该字段未记录使用;如共享搜索记录另有依据,按其来源单独展示,不推断全部未使用。
  • 文档给出新一轮工作的接入顺序:保存检索快照、登记轮次/候选、引用快照与知识条目、导入该轮搜索记录、追加决定及结果。沿用不可变 capture、版本化 optimization log 和 import-search 自身的更新/重复导入规则,不覆盖初始输入。
  • 对历史记录提供显式补充关联的方法;只使用已有且可核对的证据。缺少 candidate、run、context 或使用记录时显示断点/未记录,不按时间相近、列表顺序、名称或总结文字猜测关系。

Dependencies and delivery boundaries

Completion criteria

  • 脱敏 fixtures 分别覆盖 knowledge search、collect --relevance 和 campaign capture-context;看板准确显示实际路径,不画出未发生的调用。
  • 覆盖 skill-only、task-ranking、显式 IDs 跳过排序、未请求排序及历史未知;显示当次 method/scope/ranking_applied 和可得计数,不从缺字段推断排序已执行。
  • 新记录可下钻查询、来源/版本、过滤范围、排序方法、可得分数/组成和最终顺序;回归对照新增观测前后的返回引用、顺序、装配资源/正文及预算结果保持一致,旧记录缺字段仍可读取,不重算历史分数。
  • 覆盖“5 项选中、2 组装配、3 项因预算省略”的场景,并展示条数限制、上下文限制、未请求装配、零命中和缺失采集各自的状态。
  • 至少两个轮次、多个检索与同候选多 run 的示例可双向导航,不串轮次、不把重复记录视为新增知识;采用后失败、无收益及 INCONCLUSIVE 结果仍可追溯。
  • 共享搜索记录覆盖本地 v1、可移植 v2、partial/missing coverage 及重复导入;保留声明与采用依据的版本语义,缺失或不支持的版本/来源明确提示,不将未知计数显示为零。
  • 空 used_refs 显示该字段的使用情况未记录,不推断全部未使用;声明采用但正文被省略、知识版本/哈希不符、缺失 artifact 或关联身份冲突时明确提示问题。
  • 已准备上下文、声明采用和有证据支持的使用关联分别展示;没有对照证据时不声称知识导致收益,也不分摊收益百分比。
  • 现有 Dashboard 两种布局、Run review、离线 HTML 与 live refresh 使用一致数据投影;中英文、键盘访问和表格回退可用,刷新后保留有效的选择/展开状态。
  • 提供数据契约与接入文档、可离线评审的合成示例及相关解析/关联/浏览器回归结果;展示边界和历史缺口清楚,不改变 campaign verdict。记录实际依赖版本、每个交付切片及剩余验收项。

Scope and constraints

Inputs and deliverables

Ownership and authorization

  • Proposed by / actual requester: 用户在 Codex 会话中明确要求提交 issue,补齐上述两部分;本次另明确授权核实评论并据此修改 issue 或回复。
  • Status: Proposed。
  • Acting contributor: Codex,需求整理、意见核查与 issue 更新。
  • Accountable parent / implementer / goal-acceptance owner: TBD,由维护者指定。
  • Accepted decision or scoped delegation: 当前授权包括需求 issue 的创建、意见核查与正文/讨论更新;实现排期、合并与最终验收由后续明确授权决定。

Completion and closure

Activity

  1. added
    kind:taskBounded delivery task with a delegated slice
    area:harnessHarness module: profile, A/B loop, quality gates, evidence reports
    area:knowledgeKnowledge line: packs, skill library, knowledge admission
    area:usabilityUsability and collaboration: issue navigation, templates, readable reports
    status:proposedWaiting for the decision named in the status block
    on Nov 12, 2026
  2. amy-why-3459 commented on Nov 12, 2026

    @amy-why-3459
    CollaboratorAuthor
  3. zengchuang-hw commented on Nov 12, 2026

    @zengchuang-hw
    Contributor

    设计复核:方向与现有证据边界一致,但实施前建议先校正当前基线描述,并明确复用的记录契约。由 Codex 代表 zengchuang-hw 回应 本次提及;以下是客观设计反馈,不代表接受首版范围、指派负责人或承诺实施。

    1. 排序入口不能只分“catalog 搜索 / skill 排序”。 已核对 main e2f4cd1e39b369f06835cdd655ad5dc9f6808918:catalog.search_knowledge 确实按查询词出现次数排序,并按条目 ID 打破同分;但 collect 已区分 skill-only 与 task-ranking,后者覆盖 package skills 和 records。显式 IDs 路径保留调用者顺序,不执行 relevance ranking。因此图中应读取当次 selection.method/scope/ranking_applied,区分“未请求排序”“显式 IDs 跳过排序”“旧记录未知”,不要把所有 collect 画成 skill 排序。正文 Current behavior 可据此更新。

    2. 已有 capture 字段应直接复用,新增部分要明确。 capture_context 已保存 request.relevance_method、selection、items、omissions、context_omissions 和字节预算;当 relevance 有值且未给 IDs 时,还会选择 task relevance。当前 collect 虽内部计算分数,却未在返回条目中暴露逐项分数/分解。因此“展示已有 method/scope”等只读投影与“在执行点新增分数/分解记录”应分别列项,不能用当前代码对旧快照重算。新增观测的回归宜对照改动前后的返回引用、顺序和装配结果,证明只增加解释而不改变检索行为。

    3. 与 task: 支持每个优化迭代的知识库命中归因统计 #344 的边界建议以共享记录为准,不另建采用判定。 task: 支持每个优化迭代的知识库命中归因统计 #344 的通用逐轮统计仍是提案,关联实现 #384 当前仍开放;它不应被描述成 main 已交付契约。建议本 issue 的展示/导航消费同一份命中、读取、声明采用、依据及结果关联,并对版本、身份和来源不匹配明确降级。task: 支持每个优化迭代的知识库命中归因统计 #344 中候选集合与 token/费用的评论仍待取舍,本 issue 已排除费用统计,不应因关联就隐式扩大范围。空 used_refs=未记录、正文装配≠模型实际采用、失败结果仍可追溯,这些约束建议保留。

    4. 交付依赖与验收分开写。 #405 当前仍开放,已通过独立代码与浏览器复审,但其 Run review/轮次导航仍是 PR 能力,不是 main 能力。实现时应核定实际依赖版本;task: Dashboard 增强:每轮改动说明、动态 workflow 与耗时、测试进度 #345 的 workflow/改动说明关联也需共用 campaign/iteration/candidate/run 身份,不由显示名称或时间相近补关系。现有“两个轮次、同候选多 run、历史断点、两布局及 live refresh”验收方向合理。

    需要维护者决定的事项仍是: 与 #344/#384 的契约归属、首版先交付哪些采集/导航切片,以及 accountable parent、implementer、验收负责人。本条没有替维护者作出这些决定,也没有认领实现、排期或合并。

    实际检查:#406 正文和全部 1 条既有评论;#344、#379、#345、#392 正文及全部现有讨论;当前 main 的 CLI、catalog、selection、collect、capture 代码;#384/#405 的最新开放状态及 #405 已完成审查。未执行本提案的新实现、测试、浏览器验收或私有 H3 历史重建。本轮是静态设计复核,不改变 RFC、campaign verdict 或历史证据。

  4. amy-why-3459 commented on Nov 13, 2026

    @amy-why-3459
    CollaboratorAuthor

    已按用户授权核对并更新正文。上述关于排序范围、capture 字段复用、共享采用记录和依赖/验收分离的建议成立;正文现已补上对应契约与回归验收项。

    有两处状态需要按 2026-11-13 的进展更新:

    1. #384 已合并。当前 main 的 campaign import-search 已提供本地 v1 / 可移植 v2 记录,本 issue 应消费这些记录并保留其声明、采用依据及覆盖状态的区别;这不代表 task: 支持每个优化迭代的知识库命中归因统计 #344 整体已验收。
    2. #405 仍开放,但按 iterations[].context_ids 展示本轮检索的 Run review 能力已随 #401 进入主线。当前实现 已展示 method/scope、返回顺序和检索详情;后续曲线/事件功能仍按 feat: Dashboard 实验性能曲线、事件轨道与证据下钻 #405 的实际依赖版本处理。

    本次静态核对基于 main fe16b29388de29708d30ea388b589355540f3172 的 CLI、catalog、collect、capture、search attribution 和 Run review,以及相关 issue/PR 的现状。只更新需求正文与本条说明,未实施新功能、运行测试或重建 H3 历史;首版范围及负责人仍待维护者确认,未改变 campaign verdict。

  5. amy-why-3459 commented on Nov 13, 2026

    @amy-why-3459
    CollaboratorAuthor
  6. zengchuang-hw commented on Nov 13, 2026

    @zengchuang-hw
    Contributor

    独立增量设计复核:当前正文已回应我上轮意见,本轮未发现需要再次修改需求正文的阻塞。

    复核对象为 2026-11-13 10:05 UTC 的正文;代码依据 main fe16b29388de29708d30ea388b589355540f3172。已独立核实检索入口、skill-only/task-ranking、显式 IDs 跳过排序、capture 现有字段及主线 Run review 展示。正文现在正确区分 #384/#401 已落地主线与 #405 尚未合入的能力,也明确保留 v1/v2 证据语义、未知计数、失败结果和非因果边界。上轮关于 #384 未合并、Run review 检索仅在 #405 分支的描述是当时状态,不再适用于当前基线。

    以下只是落实已写明验收项时的实现提示,不是新需求或新的批准门槛:

    • 同候选多 run:当前 _round_candidates 读取单一 run_id,measured 遇到多结果会标为 ambiguous;实现正文已有的多 run 导航时,要保留每次运行身份和失败结果,不能取第一条或最新一条替代。
    • 新排序观测:knowledge_capture._item 和 optimization_story._capture 都有字段投影边界;落实已要求的兼容扩展时,需要让新增解释字段贯穿执行点、快照和展示,同时保持原返回顺序与装配结果。
    • 脱敏与可信度:沿用正文已有的脱敏约束;资源/快照 hash 一致不证明模型已收到或采用内容。公开投影仍须检查 query、路径和片段,不能因为状态为 verified 就自动公开。

    建议沿正文已有的“现有字段展示 → 缺失观测采集 → 共享记录关联导航”分片推进。与 #344 的契约归属、首版取舍及负责人仍由维护者决定;本条不代定负责人、不认领实现,也不表示整体验收通过。

    实际检查:独立静态阅读当前正文、上述主线源码、#344/#345 正文及仓库协作/隐私约束;核对 #405 为开放未合入(97b5dad06d0bb792455e5da44d9c668b3f004c97)。本轮没有参考他人的检视意见,没有运行尚未交付的新实现、浏览器验收或重建私有 H3 历史。

  7. Dong1017 commented on Nov 17, 2026

    @Dong1017
    Collaborator

    Posted by an agent on behalf of @Dong1017 (incremental only; points already raised by others or decided by the owner are not repeated).

    Incremental note: the delivery for this issue has already merged, but the status block still says Proposed and asks for a first-slice and owner decision.

    1. #428 ("补齐知识检索解释与逐轮候选证据导航", Refs #406) was merged on 2026-11-14 as 57f8e8a4, approved by @zengchuang-hw and cursor[bot]. Its body says it covers Scope A/B: execution-time observation, read-only display, shared-record integration and navigation. #405 was also merged on 2026-11-13 (d61552eb). The body still says "feat: Dashboard 实验性能曲线、事件轨道与证据下钻 #405 在本次核对时仍开放", and no thread comment links feat: 补齐知识检索解释与逐轮候选证据导航 #428.
    2. Under docs/agent-collaboration.md:226-227, status:proposed stays until the decision owner records a decision. The work has now landed without that record. The maintainers need to record two things here: (a) acceptance of scope and owner, retroactive to feat: 补齐知识检索解释与逐轮候选证据导航 #428; (b) whether feat: 补齐知识检索解释与逐轮候选证据导航 #428 completes the issue or leaves items for task: 支持每个优化迭代的知识库命中归因统计 #344. feat: 补齐知识检索解释与逐轮候选证据导航 #428 itself lists its non-goals: the full non-returned candidate set, token/cost statistics, and causal attribution.

    Scope: PR state, merge commit and approvals via the GitHub API, plus the issue body and comments. I did not re-review the #428 diff against the acceptance items.

  8. zengchuang-hw commented on Nov 17, 2026

    @zengchuang-hw
    Contributor

    Independent implementation checkpoint for the already-closed #406. Reviewed current main 61303fa8150472dfbda62ec8d77908bc05c73f3b; this does not reopen the Issue or restart any campaign.

    This is new implementation/test evidence relative to my previous design-only checkpoint at fe16b29388de29708d30ea388b589355540f3172 and initial design review. I inspected current code and ran focused tests first, then checked only my own earlier replies for duplication. I did not use other reviewers' conclusions or the amended Issue narrative as acceptance criteria.

    Now independently verified in the current implementation

    • src/inferrsi/knowledge/retrieval.py:observation records entrypoint/calls, request, selection, returned items, counts and assembly state. src/inferrsi/apps/views/retrieval_projection.py:retrieval_view preserves recorded explanations and distinguishes missing/unsupported history; legacy gaps do not turn into invented rankings, scores or assembly.
    • The execution/ordering fixtures cover three ranking methods, score decomposition, tie ordering, exact material/context budgets and omissions. The projected observation's item order matches the actual returned item order. Catalog observation is also tested not to mutate records or returned ordering.
    • read_searches consumes existing versioned shared search records, verifies capture/revision/passage and candidate/run/source associations, and exposes explicit mismatch states. project_knowledge keeps same-candidate multiple runs separate and marks conflicting rounds/duplicate runs rather than selecting a convenient result. Synthetic navigation retains both FAIL and INCONCLUSIVE outcomes; incomplete shared counts stay null instead of becoming zero.
    • Public projection sanitizes even a capture labelled verified. Tests cover secret/private-path redaction, malformed/duplicate-key/symlinked records, mismatched declaration hashes and invalid declaration identities without breaking dashboard rendering. Association states remain recorded relationships, not causal performance attribution or proof of model adoption.

    Checks actually run

    pytest -o addopts='' -q tests/test_retrieval_projection.py tests/test_optimization_trace.py
    → 81 passed, 0 skipped in an isolated checkout of the pinned main revision.

    These checks provide implementation evidence for the earlier design concerns around explanation propagation, shared-record consumption, multi-run identities, unknown history and redaction. They are not an exhaustive privacy proof or a causal-attribution experiment.

    Still outside this checkpoint

    No real-browser click/live-refresh test, installed-wheel verification, full suite, private-history reconstruction or hardware/model experiment was run. The original-version/current-governing-schema gap in the clean requirements remains unresolved; I am not reconstructing amended acceptance criteria from discussions or treating closed state as acceptance evidence. Detailed UI completion and any broader contract acceptance therefore remain unverified by this review.

    No defect surfaced in this bounded CPU implementation check. No source, historical evidence, verdict, ranking algorithm, adoption judgment, Issue state or campaign state was changed; this comment records the new verification only.

  9. zengchuang-hw commented on Nov 17, 2026

    @zengchuang-hw
    Contributor

    Independent historical-request follow-through for:

    Issue406 remains closed. My earlier inspection does not establish explicit coverage of these invitations merely by date or similar subject. This response uses the isolated original goals and current actual code, without adopting source-comment merger/approval or retrospective-acceptance assertions.

    Current independently checked scope: main 0311393c62330c041a01238dc86823b80e020561. In apps/views/retrieval_projection.py, retrieval_view uses the recorded observation when structurally supported; otherwise it labels entrypoint/candidate details as unknown or not recorded instead of replaying today's ranking. Explicit IDs, ranking not requested, and unknown ranking are separate states. Imported search validation distinguishes v1/v2 and requires additional v2 corpus/declaration identity. Partial coverage cannot carry complete counts. project_knowledge links explicit context/round/candidate/run records, retaining missing, ambiguous and round-conflict states.

    Direct answer to the administrative boundary question: this evidence supports the existence and selected behavior of a read-only association/presentation implementation. It does not prove that a responsible acceptance owner, remaining-work allocation against #344, or an authoritative delivery-completion record has been recorded. I have not independently authenticated those administrative decisions, and neither the closed flag nor a test pass fills that gap. Any completion record should separately identify the delivered display/navigation slice, the accountable acceptance owner, and work retained under #344; this comment does not assign those responsibilities or change their scope. In particular, navigation from a retrieval to an execution is not causal attribution, corpus exhaustiveness, token/cost accounting or automatic knowledge promotion.

    Actual fresh checks: on that exact main, test_retrieval_projection.py, test_optimization_story.py, and test_dashboard_live.py completed 106 passed, one pytest-plugin deprecation warning, in 17.67s. These exercise CPU projection/story/server-side live behavior. They are not browser interaction acceptance or confirmation of historical administrative records.

    Still unverified: exact original immutable acceptance wording, authoritative owner/status reconciliation, two-layout keyboard/locale behavior, and browser offline/live selection-refresh behavior were not fully revalidated here. No new browser run, complete suite, current-source reconstruction, candidate generation, H3 rerun or device/model experiment was performed. Missing historical scores, counts and relationships remain unknown; I have not guessed them from present algorithms or names/timestamps.

    AGENTS/review guidance was checked. No tracker, code, evidence, validator or Issue status was changed; no reopening, merge, paid execution, promotion or stopped-work restart occurred. This is bounded verified follow-through, not whole-Issue or parent-goal PASS.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    area:harnessHarness module: profile, A/B loop, quality gates, evidence reportsarea:knowledgeKnowledge line: packs, skill library, knowledge admissionarea:usabilityUsability and collaboration: issue navigation, templates, readable reportskind:taskBounded delivery task with a delegated slicestatus:proposedWaiting for the decision named in the status block

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions