数据截至Data as of
2026-10-21T05:07Z
RGS-N132D：把 XMM-Newton RGS 的“空间”变成“光谱”RGS-N132D: turning XMM-Newton RGS space into spectrum
项目对读者的名字是 RGS-N132D（代码与文件夹名 S2Fitting，PI 于 2026-10-20 决定，D32）：一条可重复的 XMM-Newton RGS 空间到光谱前向拟合管线。SAS（Science Analysis System）负责仪器校准、事件、抽谱区域、光谱与响应矩阵；Python 负责配置、编排、验证、Sherpa 模型与结果发布。当前科学目标（PI 于 2026-10-19/20 设定，PI 第 29 号决定及其第 33、34 号修订）是：用一条合并窄带（核心 14.69-23.45 Å，先做纯氧阶段 18.20-23.45 Å）同时建模 O VII、O VIII、N VII 与 Fe L，得到发射气体的空间分布与速度分布。本页是工作台：这个项目是什么、现在走到哪里、什么还没解决、下一步做什么。The project's reader-facing name is RGS-N132D (code and folder name S2Fitting, decided by the PI on 2026-10-20, D32): a reproducible XMM-Newton RGS spatial-to-spectral forward-fitting pipeline. SAS owns instrument calibration, events, extraction regions, spectra and response matrices; Python owns configuration, orchestration, validation, Sherpa model composition and result publication. The current science goal (set by the PI on 2026-10-19/20, PI decision 29 as revised by decisions 33 and 34) is to model O VII, O VIII, N VII and Fe L together in one combined narrow band (core 14.69-23.45 Angstrom, an oxygen-only stage at 18.20-23.45 Angstrom first) and obtain the emitting gas's spatial and velocity distributions. This page is the workbench: what this is, where it stands, what is unsolved, and what comes next.
本次更新What's new
2026-10-20 刷新 12026-10-20, refresh 1
在跑In flight
3 项（第 34 号实验、第 33 号实验 重启、红队第 30 号）3 jobs (experiment 34, the experiment 33 restart, red team 30)
组会（最新一天）Group meeting (latest day)
2026-10-21（UTC，进行中） —— 成果：Suzuki 等人 (2020) 全文深读完成（LIT4 双 agent，未复核）——该文已做积分三联诊断、三温度 NEI 与按 roll/能量分段的 Chandra 轮廓响应平滑，因此“首次 RGS 谱/扩展源改正/f-r 测量”都不是本项目的创新点；唯一明确的空间新结果是 O VII 禁线天空偏移（约 3.2-3.6″ 北向），而 \(\sigma_v \approx 500\) km/s 与 Suzuki 的 438/445 km/s 属同一量级而非新发现。PI 第 36 号决定：今后每条 headline 必须带“对比 Suzuki 2020”的创新性标签。GOF 共因红队（红队第 32 号，未复核）：泊松模拟本身正确；18 个廉价自由参数（归一 + 背景尺度）只买到 \(\Delta C = -17.4\) / \(-49.7\)，超出仍为 \(z = +7.13\) / \(+10.16\)；O VIII 失配集中在 Ly-alpha 翼（占 45.1 % 的 \(\chi^2\)）加蓝边超出，O VII 失配平滑弥散。分阶段 roll 拟合第 1 阶段（第 36 号实验，未复核）：三观测 L 模型过新收敛门，\(v_0 = 34.5\) km/s、\(|g| = 121.7\) km/s/′、\(\sigma_v = 396.4\) km/s；L2ion 未过门。问题：机器长期饱和（CPU 97-99.9 %）；本 agent 上一会话因“响应过大”中断，本会话为全新接手。待 PI：是否把 LIT4 的某条候选 headline 提升为论文/幻灯片主结论。2026-10-21 (UTC, in progress) — Results: the full deep read of Suzuki et al. (2020) is complete (LIT4 dual agents, unreviewed) — that paper already did integrated triplet diagnostics, a three-temperature NEI model and Chandra-profile response smoothing by roll angle and energy interval, so "first RGS spectrum / extended-source correction / f-versus-r measurement" are not this project's novelty; the one clearly new spatial result is the O VII forbidden-line sky offset (about 3.2-3.6 arcsec north), while \(\sigma_v \approx 500\) km/s is the same scale as Suzuki's 438/445 km/s rather than a new finding. PI decision 36: every headline must from now on carry a "versus Suzuki 2020" novelty label. The GOF common-cause red team (red team 32, unreviewed): the Poisson simulation is correct; 18 cheap free parameters (normalisations plus background scales) buy only \(\Delta C = -17.4\) / \(-49.7\), leaving the excess at \(z = +7.13\) / \(+10.16\); O VIII's misfit concentrates in the Ly-alpha wings (45.1 % of the \(\chi^2\)) plus a blue-edge excess, while O VII's is smooth and distributed. Stage 1 of the staged roll-angle fit (experiment 36, unreviewed): the three-observation L model passes the new convergence gate with \(v_0 = 34.5\) km/s, \(|g| = 121.7\) km/s per arcmin and \(\sigma_v = 396.4\) km/s; L2ion does not pass. Problems: the machine has been saturated (CPU 97-99.9 %); this agent's previous session died ("response too large") and this is a fresh session taking over. For the PI: whether to promote one of LIT4's candidate headlines to the paper's or the deck's central claim.
更早的天（2026-10-20：14 观测联合拟合第一批结果与 GOF 共因排除；2026-10-19：PI 第 29 号决定设定目标；handbook 首建与评审 1 返工；11 个新观测约简完成；NIM 代理停摆 3 小时）见组会记录。Earlier days (2026-10-20: the 14-observation joint fit's first batch and the GOF common-cause exclusions; 2026-10-19: the goal set by PI decision 29; the handbook's first build and review-1 rework; 11 new observations reduced; the NIM proxy down for 3 hours) are in the group-meeting record.
本页目录On this page
管线骨架Pipeline skeleton
来源：docs/architecture.md（数据流与所有权边界）、cognition-os/experiments/E018_real-data-readiness.md（前向模型与讨厌参数组）。Source: docs/architecture.md (data flow and ownership boundaries), cognition-os/experiments/E018_real-data-readiness.md (forward model and nuisance set).
目标表Objectives
| 目标Objective | 状态Status | 当前最佳数值Best number now | 来源Source | 趋势Trend |
|---|---|---|---|---|
| 合并氧带（O VII+O VIII+N VII+Fe L）的速度场Velocity field of the combined oxygen band (O VII + O VIII + N VII + Fe L) | ProvisionalProvisional | 14 obs 第一批（未复核，PI 第 35 号决定后仅作参考）： O VII \(v_0 = 41.6 \pm 9.1\) km/s \(|g| = 137.4 \pm 11.3\) km/s/′ \(\sigma_v = 466.7 \pm 10.6\) km/s O VIII \(|g| = 0\)（边界）、\(\sigma_v = 497.8 \pm 6.6\) km/s 分阶段三观测（第 36 号实验，未复核）： \(v_0 = 34.5\)、\(|g| = 121.7\) km/s/′ \(\psi = -38.0^\circ\)、\(\sigma_v = 396.4\) km/s14-obs first batch (unreviewed, reference-only after PI decision 35): O VII \(v_0 = 41.6 \pm 9.1\) km/s \(|g| = 137.4 \pm 11.3\) km/s/arcmin \(\sigma_v = 466.7 \pm 10.6\) km/s O VIII \(|g| = 0\) (bound), \(\sigma_v = 497.8 \pm 6.6\) km/s staged three-observation (experiment 36, unreviewed): \(v_0 = 34.5\), \(|g| = 121.7\) km/s/arcmin \(\psi = -38.0^\circ\), \(\sigma_v = 396.4\) km/s |
E033 报告report（未复核，仅参考）(unreviewed, reference-only) E036 报告report（未复核）(unreviewed) |
\(\sigma_v\) 仍在 400-500 km/s；分阶段拟合取代 14 观测一步跳\(\sigma_v\) still 400-500 km/s; the staged fit replaces the 14-observation jump |
| \(\sigma_v\) 是内部弥散还是整体分裂？Is \(\sigma_v\) internal dispersion or a bulk split? | ProvisionalProvisional | \(\Delta\mathit{Cstat} = -151.7\)（7 参数；红队第 29 号 复算） 分裂法线离 O1/O2 色散法线 \(1.4\)-\(2.5^\circ\)\(\Delta\mathit{Cstat} = -151.7\) (7 parameters; red team 29 replay) split normal within \(1.4\)-\(2.5^\circ\) of the O1/O2 dispersion normals |
I029, RT29（红队第 29 号 未复核）(red team 29 unreviewed) |
不是泊松涨落，但与色散轴简并，需新 roll 角not a Poisson fluke, but degenerate with the dispersion axes; needs new roll angles |
| 合并带的边与谱线清单（奥卡姆）The combined band's edges and Occam line list | ProvisionalProvisional | 核心 14.69-23.45 Å；纯氧 18.20-23.45 Å；8 个混合成分，每个 \(\Delta\mathit{BIC} > 0\) 且 Asimov 可恢复core 14.69-23.45 Angstrom; oxygen-only 18.20-23.45 Angstrom; 8 blends, each \(\Delta\mathit{BIC} > 0\) and Asimov-recoverable | E032 报告report（未复核）(unreviewed) |
已定稿，待复核fixed, awaiting review |
| f 线天空偏移的形态：刚性平移还是比值梯度？The f-line sky offset's morphology: rigid shift or ratio gradient? | 改读re-read | 梯度 \(\Delta C = -75.33\) 优于平移 \(-53.45\)（同 2 参数）；给定梯度后平移只剩 \(\Delta C = -3.67\)（吸收 93 %）the gradient's \(\Delta C = -75.33\) beats the shift's \(-53.45\) (both 2 parameters); given the gradient, the shift retains only \(\Delta C = -3.67\) (93 % absorbed) | RT30 报告report（未复核）(unreviewed) |
“f/r 分布不同”高置信；“刚性平移”不再是唯一读法high confidence the f/r distributions differ; a rigid shift is no longer the unique reading |
| 可用数据Usable data | 就绪ready | 14 obs（3 + 11）；28 个合并带响应包（875 再生 + 105 复用），4443 文件哈希通过；26/28 单元可用14 obs (3 + 11); 28 combined-band response packs (875 regenerated + 105 reused), 4443 files hash-verified; 26 of 28 units usable | E024、E034 |
已就绪（第 34 号实验 完成）ready (experiment 34 complete) |
| 仪器系统误差Instrument systematics | 基本关闭largely closed | \(\le 0.1\) SD（典型） \(w \approx -3\,\%\) 未解释\(\le 0.1\) SD (typical) \(w \approx -3\,\%\) unexplained |
I028、I025 |
讨厌参数组有效the nuisance set works |
| 结果相对 Suzuki 2020 的创新性The results' novelty relative to Suzuki 2020 | 已建立规则rule set | “新”的只有：O VII 禁线天空偏移（约 3.2-3.6″ 北向）；O VII 空间速度梯度（方向不稳健） “同一量级”的：\(\sigma_v \approx 500\) km/s（Suzuki 438/445）；全球 f/r；线清单；用 Chandra 轮廓改响应 来源：LIT4 双报告 §4/§6（PI 第 36 号决定）"New" only: the O VII forbidden-line sky offset (about 3.2-3.6 arcsec north); a spatially resolved O VII velocity gradient (direction not robust) "Same scale": \(\sigma_v \approx 500\) km/s (Suzuki 438/445); the global f/r; the line inventory; using Chandra-profile response convolution Source: the two LIT4 reports, sections 4 and 6 (PI decision 36) |
LIT4 报告reports（未复核）(unreviewed) |
每条 headline 今后必须带创新性标签every headline must now carry a novelty label |
来源：cognition-os/02_CURRENT_QUESTION.md（子问题表）、cognition-os/03_TRUST_TABLE.md（每条结论的当前信任标签）、cognition-os/04_DECISION_LOG.md（D29，PI 短期目标）。SD 指 Fisher 标准误（standard deviation）。Source: cognition-os/02_CURRENT_QUESTION.md (sub-question table), cognition-os/03_TRUST_TABLE.md (the current trust label of each claim), cognition-os/04_DECISION_LOG.md (D29, the PI's short-term goal). SD means Fisher standard deviation.
未解问题Open problems
| 问题Problem | 为何重要Why it matters | 最佳假设Best hypothesis | 下一个检验Next test | 状态Status |
|---|---|---|---|---|
| 两个窄带都不通过泊松拟合优度检验（O VII z \(\approx +8\)、O VIII \(\approx +11\)）Both narrow bands fail the Poisson goodness-of-fit test (O VII z \(\approx +8\), O VIII \(\approx +11\)) | 所有速度数值都以速度模型为条件；这是当前主要限制Every velocity number is conditional on the velocity model; this is the current main limitation | 已排除：多余谱线、背景、LSF 宽度、Ly-alpha 形状（第 28 号实验/第 29 号实验）、粗模板失配（第 31 号实验 分块到 4x4 仍 \(z > +3\)）、非高斯速度轮廓与 LSF 翼（第 35 号实验 中只有“高斯+宽高斯”同时改善两带，\(\sigma_2 \approx 750\)-920 km/s，占 \(\approx 0.36\)，但 GOF 仍不合格）、逐观测 x 条带归一与背景尺度（红队第 32 号：18 个参数只买 \(\Delta C = -17.4\) / \(-49.7\)，\(z\) 仍 \(+7.13\) / \(+10.16\)）。剩余失配是平滑、跨带、带边加权的超出，加（仅 O VIII）Ly-alpha 宽度缺陷Ruled out: extra lines, background, LSF width, Ly-alpha shape (experiment 28/experiment 29); coarse template mismatch (experiment 31, still \(z > +3\) up to 4x4 blocks); non-Gaussian velocity profiles and LSF wings (in experiment 35 only Gaussian-plus-broad-Gaussian improves both bands, \(\sigma_2 \approx 750\)-920 km/s at fraction \(\approx 0.36\), but the GOF still fails); per-observation x strip normalisations and background scales (red team 32: 18 parameters buy only \(\Delta C = -17.4\) / \(-49.7\), leaving \(z\) at \(+7.13\) / \(+10.16\)). What remains is a smooth, band-wide, band-edge-weighted excess plus (O VIII only) a Ly-alpha width defect | 逐观测背景光滑形状自由度（第 37 号实验，在 lang2 上，进行中）；若仍不合格则问题在源模型（连续谱模板或线宽）a per-observation smooth background shape freedom (experiment 37, on lang2, in progress); if it still fails, the defect is in the source model (continuum template or line width) | 在跑running |
| f 线偏移是刚性平移还是 f/r 比值梯度Is the f-line offset a rigid shift or an f/r ratio gradient | 它决定梯度读数的系统误差（0.3-0.6 SD）与物理图像（位置不同 vs 比例不同）It sets the gradient reading's systematic (0.3-0.6 SD) and the physical picture (displaced gas versus a different ratio) | 22.19 Å 混合线被拒（红队第 30 号：0 % 吸收，几何反号）；比值梯度拟合更好并吸收 93 % 的平移 \(\Delta C\)the 22.19 Angstrom blend is rejected (red team 30: 0 % absorbed, geometry sign-flipped); the ratio gradient fits better and absorbs 93 % of the shift's \(\Delta C\) | 14 观测堆叠上同时放开两种形态（待 PI）free both morphologies on the 14-observation stack (for the PI) | 待 PI 决定for the PI |
| 三观测双成分分裂与色散轴简并The three-observation two-component split is degenerate with the dispersion axes | 分裂增益 \(\Delta\mathit{Cstat} = -151.7\) 不是泊松涨落，但 O1/O2 色散法线只差 \(3.9^\circ\)，分裂法线落在其 \(1.4\)-\(2.5^\circ\) 内the split's \(\Delta\mathit{Cstat} = -151.7\) is not a Poisson fluke, but the O1/O2 dispersion normals differ by only \(3.9^\circ\) and the split normal lies within \(1.4\)-\(2.5^\circ\) of them | 天空结构 vs 锁轴仪器效应当前不可分（红队第 29 号）sky structure versus an axis-locked instrumental effect cannot currently be separated (red team 29) | 分阶段加观测（第 36 号实验）：S3 三观测 L 已过门，逐级加 roll 角到 14（S4-S10 待跑，L2ion 未收敛）the staged observation additions (experiment 36): S3's three-observation L passes the gate, roll angles are added stepwise to 14 (S4-S10 pending, L2ion unconverged) | 在跑running |
| 交叉色散尺度 \(w \approx -3\,\%\) 无法用任何已测响应误差解释The cross-dispersion scale \(w\) of about \(-3\,\%\) is explained by no measured response error | \(w\) 与模板宽度简并，直接影响 \(\sigma_v\) 的物理读法\(w\) is degenerate with the template width, which directly affects the physical reading of \(\sigma_v\) | 条带边 \(0.5′\) 内的接受度误差，或模板底座A near-edge acceptance error within \(0.5′\) of the strip edge, or a template pedestal | 合并带拟合中继续带着 \(w\) 自由keep \(w\) free in the combined-band fits | 开着open |
| 分阶段拟合的收敛门与双离子模型The staged fit's convergence gate and the two-ion model | dispatcher 更正 1 把“优化器状态”换成“起点一致 + 有限差分”门后，L 过门、L2ion（双离子）8 个起点只有 2 个一致仍不过——两离子是否该分开拟合尚无定论after dispatcher correction 1 replaced the "optimizer status" gate with "start agreement plus finite difference", L passes but L2ion (two-ion) still fails with only 2 of 8 starts agreeing — whether the two ions should be fitted separately is undecided | 按更正 2：一个模型失败不停整个 job，L2ion 记为未收敛，继续 S4-S10per correction 2: one model's failure does not stop the job, L2ion is recorded as unconverged, and S4-S10 continue | 继续 S4-S10 阶段（每级过门后加下一个 roll 角）continue stages S4-S10 (add the next roll angle once each stage passes the gate) | 未复核unreviewed |
来源：cognition-os/02_CURRENT_QUESTION.md（第 16-23 行）、cognition-os/08_RUNNING.md（在飞的 第 18c 号实验/第 20 号实验/第 21 号实验/RT22/RT23）、cognition-os/REDTEAM_QUEUE.md（RT22、RT23 行）、cognition-os/insights/I029_n132d-ovii-velocity-stage1.md（模型不足与下一步）。Source: cognition-os/02_CURRENT_QUESTION.md (lines 16-23), cognition-os/08_RUNNING.md (the in-flight experiment 18c/experiment 20/experiment 21/RT22/RT23), cognition-os/REDTEAM_QUEUE.md (the RT22 and RT23 rows), cognition-os/insights/I029_n132d-ovii-velocity-stage1.md (model inadequacy and next steps).
接下来 1-4 周Next 1-4 weeks
- 继续 第 36 号实验 的分阶段拟合（S4-S10）：每级通过收敛门后加下一个 roll 角观测；L2ion 未收敛按 dispatcher 更正 2 如实记录。目标：\(\sigma_v\) 与梯度随观测数的稳定性；输入
.agent/jobs/2026-10-20-e036-staged-rolls-lang2/REPORT.luna.md；方法：按 PI 第 35 号决定在 lang2 上跑；成功：每级过门且参数漂移有解释；失败：记录未收敛并回退上一级；预计 2-3 天。先读：该 job 的BRIEF.md与两条CORRECTION_*.md。Continue experiment 36's staged fit (S4-S10): add the next roll-angle observation once each stage passes the convergence gate; record L2ion's non-convergence openly per dispatcher correction 2. Goal: the stability of \(\sigma_v\) and the gradient with observation count; input.agent/jobs/2026-10-20-e036-staged-rolls-lang2/REPORT.luna.md; method: run on lang2 per PI decision 35; success: each stage passes the gate and parameter drifts are explained; failure: record the non-convergence and fall back one stage; about 2-3 days. Read first: that job'sBRIEF.mdand the twoCORRECTION_*.mdfiles. - 等 第 37 号实验（背景形状自由度）回报后复核其 GOF 结论；若 \(z < +3\) 达成则“背景形状”确认为 GOF 共因，否则把平滑带边超出归给源模型（连续谱模板或线宽）。输入
.agent/jobs/2026-10-21-e037-background-shape/。Once experiment 37 (background shape freedom) reports back, triage its GOF conclusion; if \(z < +3\) is reached, "background shape" is confirmed as the GOF common cause, otherwise the smooth band-edge excess is attributed to the source model (continuum template or line width). Input:.agent/jobs/2026-10-21-e037-background-shape/. - 把 第 24 号实验–第 37 号实验、V001、红队第 27 号–红队第 33 号、LIT1-LIT4 记入 cognition-os（信任表、实验页、REDTEAM_QUEUE）；f 线读法按 红队第 30 号 更新 第 29 号洞察 措辞；headline 一律带“对比 Suzuki 2020”创新性标签（PI 第 36 号决定）。Record experiment 24-experiment 37, V001, red team 27-red team 33 and LIT1-LIT4 into cognition-os (trust table, experiment pages, REDTEAM_QUEUE); update the insight 29 f-line wording per red team 30; every headline carries a "versus Suzuki 2020" novelty label (PI decision 36).
- f 线判别检验（待 PI）：14 观测堆叠上同时放开刚性平移与比值梯度，用 roll 角杠杆分开两种形态；红队第 33 号（进行中）先回答我们的偏移与 Suzuki 2020 的 f/r 图是否相容。The f-line discriminant (for the PI): free both the rigid shift and the ratio gradient on the 14-observation stack and use the roll-angle leverage to separate them; red team 33 (in progress) first answers whether our offset is compatible with Suzuki 2020's f/r maps.
- Fe L 阶段：纯氧阶段稳定后，把 Fe L 加入合并带（PI 第 33 号决定的分阶段精神），测 Fe L 速度场与氧对比。The Fe L stage: once the oxygen-only stage is stable, add Fe L to the combined band (the staging spirit of PI decision 33) and measure the Fe L velocity field against oxygen.
来源：cognition-os/08_RUNNING.md 第 5、7 节（待派任务与 CoPI 建议）与 cognition-os/02_CURRENT_QUESTION.md（下一步行动）。第 3 条待 PI 决定；第 5 条是 CoPI 的建议方向。Source: cognition-os/08_RUNNING.md sections 5 and 7 (ready-to-dispatch tasks and CoPI proposals) and cognition-os/02_CURRENT_QUESTION.md (next concrete actions). Item 3 awaits the PI; item 5 is a CoPI proposal.
本次更新（2026-10-21，刷新 1 续）What's new (2026-10-21, refresh 1 continued)
- 组会页新增 2026-10-21（进行中）一天：Suzuki 2020 深读、GOF 共因红队、分阶段 roll 拟合第 1 阶段、第 37 号实验 与 红队第 33 号 在跑。The group-meetings page gains the (in-progress) day 2026-10-21: the Suzuki 2020 deep read, the GOF common-cause red team, stage 1 of the staged roll-angle fit, and experiment 37 and red team 33 in flight.
- 目标表新增“结果相对 Suzuki 2020 的创新性”一行（PI 第 36 号决定；“新”的只有 f 线偏移与空间梯度，\(\sigma_v\) 量级与 Suzuki 一致）。The objectives table gains a "novelty relative to Suzuki 2020" row (PI decision 36; only the f-line offset and the spatial gradient are new, while the \(\sigma_v\) scale matches Suzuki's).
- GOF 未解问题行更新：红队第 32 号 排除了归一与背景尺度（18 参数只买 \(\Delta C = -17.4\) / \(-49.7\)）；下一个检验是 第 37 号实验 的背景形状自由度。The GOF open-problem row is updated: red team 32 ruled out normalisations and background scales (18 parameters buy only \(\Delta C = -17.4\) / \(-49.7\)); the next test is experiment 37's background shape freedom.
- 14 观测一步跳的读法按 PI 第 35 号决定降为“仅参考”，主线改为 第 36 号实验 的分阶段拟合（S3 L 已过门：\(v_0 = 34.5\)、\(|g| = 121.7\) km/s/′、\(\sigma_v = 396.4\) km/s）。The 14-observation jump is demoted to reference-only per PI decision 35; the main line is now experiment 36's staged fit (S3 L passes the gate: \(v_0 = 34.5\), \(|g| = 121.7\) km/s per arcmin, \(\sigma_v = 396.4\) km/s).
- 计划表换血：分阶段拟合续跑、第 37 号实验 复核、cognition-os 记录范围扩到 第 37 号实验/红队第 33 号/LIT4。The plan is refreshed: continue the staged fit, triage experiment 37, and extend the cognition-os recording scope to experiment 37 / red team 33 / LIT4.
刷新 1 的前半（更名 RGS-N132D、组会页首建、2026-10-20 各行）见本页 2026-10-20 的更新条目与 .agent/jobs/2026-10-19-handbook-build/。未复核的任务一律标“未复核”，分诊后以信任表为准。For the first half of refresh 1 (the RGS-N132D rename, the group-meetings page, the 2026-10-20 rows) see this page's 2026-10-20 update entry and .agent/jobs/2026-10-19-handbook-build/. Anything unreviewed is labelled as such; after triage the trust table governs.