Skip to content

Feature roadmap 2026-11: CI action, standalone audit, assertion-execution gate, lite track #69

New issue
New issue
Open
Open
Feature roadmap 2026-11: CI action, standalone audit, assertion-execution gate, lite track#69

Description

@surlymochan
surlymochan
opened on Nov 8, 2026

Feature roadmap (2026-11-08). Direction set by the owner: iteration is feature-first; coverage top-ups, dead-code and doc-drift work are filler only. Full doc: goals workspace FEATURE-ROADMAP.md. Work items, in priority order:

Each item ships as its own PR(s) with tests; check the box when merged.

Activity

  1. surlymochan
    added a commit that references this issue on Nov 8, 2026

    feat(action): verify proof seals in CI, fail check on replay mismatch (…

    e2bfc01

  2. surlymochan commented on Nov 8, 2026

    @surlymochan
    surlymochan
    on Nov 8, 2026
    OwnerAuthor

    M2 tracking issue: #78 (xflow audit standalone mode). Sub-issue attach via API returned 404 (sub-issues not enabled on this repo), so linkage is by reference here and in #78's body. Acceptance evidence will land on #78.

  3. surlymochan
    added a commit that references this issue on Nov 9, 2026

    M3 slice 4: acceptance demo for the I6d assertion-execution gate (#69) (

    7b39c2c

  4. surlymochan commented on Nov 9, 2026

    @surlymochan
    surlymochan
    on Nov 9, 2026
    OwnerAuthor

    M3 验收演示（slice 4，PR #90，已合并 7b39c2c）：正负双对照、端到端证明门禁。

    演示方式

    scripts/demo-i6d-gate.sh（已进 main）走真实 xflow audit CLI，不走 JS API：在临时演示仓（自带 I6d 插件）里对两个 diff 各跑一次审计。演示本身由 test/i6d-acceptance-demo.test.js 锁定进套件。

    RED 对照——含「从不执行断言」的 diff 应判红

    diff 新增一个回调从不被调用的测试（assert value == 42 永不执行），声称 pytest test_dead.py — 2 tests, all pass：

    xflow audit: mismatch
- ✓ pytest test_dead.py (claimed exit 0, 2 passed, 0 failed; replay exit 0, pass 2)
i6d: ✗ unexecuted_assertions (1/3 executed, 1 unexecuted)
  unexecuted: test_dead.py:3: assert value == 42
audit exit=1 verdict=mismatch i6d=unexecuted_assertions

    回放本身全绿（2 passed），是门禁、不是回放，把它翻红。

    GREEN 对照——正常 diff 应判绿

    diff 只新增真实执行的断言，同一声称：

    xflow audit: verified
- ✓ pytest test_dead.py (claimed exit 0, 2 passed, 0 failed; replay exit 0, pass 2)
i6d: pass (1/2 assertions executed, 0 skipped, 1 out of scope)
audit exit=0 verdict=verified i6d=pass

    （1 out of scope 是基线里不在 diff 内的旧断言，定域行为符合预期。）

    结论

    本次演示完整证明 M3 验收标准：门禁证明每条新增/改动断言在绿跑中真实执行过，未执行的断言不计绿——RED 例中断言从未执行、测试却全绿，门禁仍判 mismatch；GREEN 例中断言全部执行，判 verified。证据：PR #90（https://github.com/surlymochan/evalid/pull/90），演示脚本 scripts/demo-i6d-gate.sh，本地全量 npm test 511 项、510 通过、0 失败、1 跳过。

    M3 复选框不由本轮勾选，请主轮核验后勾选。

  5. surlymochan commented on Nov 9, 2026

    @surlymochan
    surlymochan
    on Nov 9, 2026
    OwnerAuthor

    复选框勾选说明（2026-11-09 15:0x 轮记账）：GitHub Action + PR 评论账本一项已完成——M1 端到端验收于 01:03 轮通过并发 v0.5.14（三项：正常 PR 绿 check＋账本评论自动生成、篡改 PR 红 check＋评论指出封印破裂点、CI 回放与本地 xflow proof 一致）；SARIF 导出（本项末句）由 #92 合并、随 v0.5.18 发布（action/sarif.mjs＋verify-proof.mjs --sarif-out＋upload-sarif 接入，端到端取证见 #73 证据评论）。故勾选本项；下一里程碑待明日首轮定（候选 lite 快速轨）。

  6. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M2 ACCEPTED (2026-11-11 09:03 round): xflow audit standalone mode — verdicts on 3 real historical PRs (senpi#2606/#4801, omo#8014) + tampered-claim case (99 claimed vs 4 replayed, exact breaking point) + forged-seal case (schema_error). Fixes G1/G2 landed in PR #117 (ff643ac). Acceptance detail: #113 comment (see issue). Known limitations for a follow-up slice: G3 bun test, G4 node_modules snapshot mismatch, prose package-scope hints, RED/baseline commit scope. Release v0.5.23 in progress (release PR).

  7. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M3 I6d 断言执行门已验收（2026-11-11 11:03 轮）：切片 1 PR #121（d15f9407）＋切片 2 PR #123（ce88be61）均已合并；验收摘要见 #119 评论 6786762925。诚实注记：3 个案例为构造的最小复现用例（非历史 bug），跑的是真实门禁代码、判定为实测。发版 v0.5.24 进行中。

  8. surlymochan
    added a commit that references this issue on Nov 11, 2026

    release: v0.5.25 (#130)

    3173f44

  9. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M6 audit 回放保真度已验收（2026-11-11 19:03 轮）：切片 1 PR #133（bun test 识别）＋切片 3 PR #136（optional 平台豁免＋bun.lock 支持，三单重跑：senpi#2606/#4801 快照翻转 consistent、omo#8014 真实 inconsistent 401 差异）均已合并；验收摘要见总单 #131 评论 6794844181。发版 v0.5.26 进行中。

  10. surlymochan
    added a commit that references this issue on Nov 11, 2026

    release: v0.5.26 (#137)

    583f9f1

  11. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M7 J1 账本 agent/model 字段落地已验收（2026-11-11 21:03 轮）：J1 写 agent/model（PR #140，afd52b32）；trend 分组两组各 n=4、手工重算逐数一致；8 真实 run 证据见 #141 评论 6797437856；验收摘要见 #138。发版 v0.5.27 进行中。

  12. surlymochan
    added a commit that references this issue on Nov 11, 2026

    release: v0.5.27 (#142) (#142)

    21d7811

  13. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M8 I6b 回放树依赖供给已验收（2026-11-12 01:03 轮）：切片 1 PR #145 合并 79bf656；切片 2 三个真实 no_test_evidence 案例重跑出真实绿判定（lite-track 22/22、workflow-loader 32/32、cli-usability 68/68，I6b 均 valid），主轮独立复现一案并重跑 audit 套件 105/102（3 失败为 pytest 不在 PATH 的既存失败）；证据 #146 评论 6803005629，验收摘要 #143。发版 v0.5.28 随后。

  14. surlymochan
    added a commit that references this issue on Nov 11, 2026

    release: v0.5.28 (#147)

    6b190c5

  15. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M9 SARIF 导出进 code scanning 已验收：6 份真实案例 SARIF 全过 schema 校验（篡改封印／声称不符各判出带确切破裂点的 finding，干净 proof 空 results）；真实仓 CI 上传已验证——code scanning analysis 2146969636（results_count 1）与 alert #3（evalid/binding-mismatch，error，open）均直读核实；验收③缺口（上传步骤 continue-on-error）已由 PR #152 修复并合并 acd7187。验收摘要见 #148 评论 6806263843，证据见 #151 评论 6806158854。随 v0.5.29 发布。

  16. surlymochan
    added 2 commits that reference this issue on Nov 11, 2026

    release: v0.5.29 (#154)

    526426b

    release: v0.5.29 (#154) (#154)

    1cee94f

  17. surlymochan commented on Nov 11, 2026

    @surlymochan
    surlymochan
    on Nov 11, 2026
    OwnerAuthor

    M10 Go/Rust 回放保真度扩展已验收（2026-11-12 05:03 轮）：go test / cargo test 声称识别＋回放内依赖供给已合（PR #157，724045b9）；4 个真实案例（testify／serde_json／hexyl）3 verified＋1 mismatch，冷缓存依赖供给均 supplied；证据 #158 评论 6808726204；验收摘要 #155（本轮主轮评论）。跟进单 #159（Go 子测试计数精度）已立，作下一里程碑种子。发版 v0.5.30 随后进行。

  18. surlymochan
    added a commit that references this issue on Nov 11, 2026

    release: v0.5.30 (#160)

    0d5d5f0

  19. surlymochan commented on Nov 12, 2026

    @surlymochan
    surlymochan
    on Nov 12, 2026
    OwnerAuthor

    M11 已验收并随 v0.5.31 发布（2026-11-12 07:03 轮）：Go test 计数精确到顶层测试。验收摘要见总单 #161 评论 6810324693；证据 #168 评论 6810293799。诚实注记：旧声称 473 系旧 parser 截断所致的幻影通过，已修正为 472；第 1 次 audit 的上游计时抖动已立跟进单 #169。

  20. surlymochan commented on Nov 12, 2026

    @surlymochan
    surlymochan
    on Nov 12, 2026
    OwnerAuthor

    M12 xflow audit --batch 整仓批量审计已验收（2026-11-12 11:03 轮收口）：三仓各 5 条已合并 PR 批量审计证据齐全（evalid 0 verified/5 mismatch、swrm 0/3/2 no_claims、senpi 重跑 2 verified/3 no_claims），判定逐条可复核。验收摘要 #171 评论 6814927635。诚实注记：0 verified 主因抽取局限与回放超时，非 PR 质量问题；#160 声称 -460 已定性为抽取 bug，跟进单 #175。发版 v0.5.32 进行中。

  21. surlymochan commented on Nov 12, 2026

    @surlymochan
    surlymochan
    on Nov 12, 2026
    OwnerAuthor

    M13 声称抽取 A/B 口径修复——已验收（13:03 轮续跑，主轮裁定）。

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

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions