Repository navigation
Retroactive sign-off for Nathan #4898
Description
Purpose
Compounding Manager owns this canonical status source for Nathan. Johannes decided that agents do not wait for Nathan. Agents self-merge through all normal gates, then ask Nathan for review. The decision removes only the wait for Nathan. @myobie is assigned to the parent and each sub-issue.
Rules
- All normal gates still apply: independent exact-head GOs, required checks, valid signed evidence, rulesets and other required human decisions.
- Agents must not use --admin or GitHub auto-merge. This tracker grants no bypass of a protected path, a signing boundary, or a risk gate.
- Nathan's earlier review holds become questions for review after merge. An item that is not merged is marked with n/a; its eventual merge SHA must be recorded here and in its sub-issue.
- A sub-issue closes only after Nathan signs off or answers its question. Silence, merge, delivery of a request and agent approval do not close a sub-issue.
- Compounding Manager updates this table with Nathan's answer and the evidence link. A design decision already made by Johannes is context, not a request to reopen that decision.
- Risk labels below describe the question and affected boundary. They do not change the repository's formal risk classification.
Status
GitHub states and merge SHAs were read on 2026-11-04. Merged change requests below have API-reported full merge SHAs. Hosted #4981 carries #4107; its merge SHA is recorded in the #4904 row. Open proposals and repository-setting questions use n/a. No merge, release, code change or CI dispatch was performed for this tracker.
GitHub limits a parent to 100 native sub-issues. This tracker has reached that limit. Later children are linked in this table and the checklist below, and each child names Parent: #4898. These body-linked children follow the same review and completion rules.
| Sub-issue | PR/issue link | Merge SHA | Question | Risk | State |
|---|---|---|---|---|---|
| #4899 | pty#209 | 74e4dbbc435b8319f96a3c5ff57dbbfb0b834795 | Does Nathan accept the merged legacy attach identity check? If accepted, will Nathan cut v0.13.0-rust.6, or name the required release change? | high | post-merge review requested |
| #4900 | fractal#4847 | 0bd919a | Does Nathan accept the merged hosted replacement for #4713, which provides Python for the web regression lane? | medium | post-merge review requested |
| #4901 | fractal#4848 | f78787b | Does Nathan accept the merged hosted replacement for #4762 and its installer ownership receipts? | medium | post-merge review requested; fixes merged: #5053/#5162 and #5054/#5163 |
| #4902 | fractal#3871 | n/a | Is the setting "Allow GitHub Actions to create and approve pull requests" disabled on purpose? If not, should Nathan enable it or provide a permitted job token? If yes, should the release job only push the branch? | medium | answered — Johannes in Nathan's place: keep the setting off; job pushes the branch only (push-only); answer |
| #4903 | fractal#4655 | n/a | Does Nathan agree with the fractal-sidebar crate boundary, seed and PR split? A yes or no is sufficient. | medium | answer requested |
| #4904 | fractal#4107 | ea2ce38 | Does Nathan accept the exact-wrapper authority proof and opt-in native-question observation contract? The previous request asked for a HIGH-risk merge window with Nathan reachable for one hour. Record any remaining operational constraint, but do not hold merge only for Nathan. #4191 must retarget to main after merge; retain the #4107 branch until then. | high | post-merge review requested |
| #4905 | fractal#3457 | n/a | Which detector fix does Nathan accept: diff against the target-branch tip, or unconditional web execution for merge_group? Will Nathan dispatch full-set Workspace CI on current main with all stories, accessibility checks and both shards? | high | answer requested |
| #4906 | fractal#1153 | n/a | Does Nathan accept the public agent.restart and error-code contract, or what change is required? | high | answer requested; review after merge |
| #4907 | fractal#1441 | n/a | Does Nathan accept the graph-first queue, read and receipt contract: immediate claim and no default local-text queue? | high | answer requested; review after merge |
| #4908 | fractal#1444 | n/a | Does Nathan accept the complete Claude and Codex task and plan producers and focused resets using the existing todo schema, or what change is required? | medium | answer requested; review after merge |
| #4909 | fractal#1464 | n/a | What disabled-read behavior and exact Shared families does Nathan accept for split A, and what retention and fingerprint contract for split B? | high | answer requested; review after merge |
| #4910 | fractal#1521 | n/a | Does Nathan accept the terminal-history web intake contract, or what change is required? | high | answer requested; review after merge |
| #4911 | fractal#1547 | n/a | Does Nathan accept a coordinated fleet upgrade with fail-closed refusal until every peer is upgraded, as in #1185, or which compatible alternative is required? | high | answer requested; review after merge |
| #4912 | fractal#1335 | n/a | Should the ordered terminal input change still depend on #937, which is now closed, or which replacement and terminal.control policy does Nathan require? | high | answer requested; review after merge |
| #4913 | fractal#1442 | n/a | Does Nathan accept the plugin-first flow with one answer-settlement owner, or what change is required? | high | answer requested; review after merge |
| #4914 | fractal#1477 | n/a | Who may publish decision cards to a person? | high | answer requested; review after merge |
| #4917 | fractal#1496 | n/a | Does Nathan accept the current-value register as the active-ask destination? | medium | answer requested; review after merge |
| #4918 | fractal#1824 | n/a | What TurnControl, queued/applied terminology and ExecutionKey reuse contract does Nathan accept? | high | answer requested; review after merge |
| #4919 | fractal#1910 | n/a | What are Nathan's answers to trace-parenting DQ1, DQ2 and DQ3 in the PR? | medium | answer requested; review after merge |
| #4920 | fractal#2242 | n/a | Does Nathan choose A, A1, A2 or B after the recorded NO-GO findings? | medium | answer requested; review after merge |
| #4921 | fractal#2370 | n/a | Does Nathan accept the runtime provenance schema and privacy contract with Chief of Staff, or what change is required? | high | answer requested; review after merge |
| #4922 | fractal#3324 | n/a | Who is the R1 maintainer, and what keyed-writer behavior does Nathan require for Source admission? | high | answered — Johannes in Nathan's place: Compounding Manager owns R1; require keyed-writer policy (manager-owns-keyed); answer |
| #4924 | fractal#3532 | n/a | What receipt-mirror format, freshness and authority contract does Nathan accept? | high | answer requested; review after merge |
| #4925 | fractal#2089 | n/a | May Compounding Manager close the draft/WIP workflow proposal because the self-merge process replaces it, or what scope must remain? | low | answer requested; review after merge |
| #4926 | fractal#2958 | n/a | May Compounding Manager close the separate-app Storybook gate because #3597 and the shared kit replace it, or what scope must remain? | low | answer requested; review after merge |
| #4927 | fractal#2471 | n/a | May Compounding Manager close the old conversation header because the workspace owns the single accessible header, or what behavior must remain? | low | answer requested; review after merge |
| #4928 | fractal#2536 | n/a | May Compounding Manager close the retired ThreadHeader layout API, or what behavior must remain? | low | answer requested; review after merge |
| #4929 | fractal#2678 | n/a | May Compounding Manager close this input-only proposal because its core landed in #2694, or what scope must remain? | low | answer requested; review after merge |
| #4930 | fractal#2818 | n/a | May Compounding Manager close the old inline Markdown PR as its author requested, or what scope must remain? | low | answer requested; review after merge |
| #4931 | fractal#3531 | n/a | Does Nathan accept assistant/mcp as a second external-assistant principal and the registry digest change? Is a release note plus a fleet upgrade, with no MCP grants until every host is upgraded, the accepted rollout? | high | answer requested; review after merge |
| #4932 | fractal#3499 | n/a | Does Nathan accept a new next-free convergence corpus with task-grant events on top of main and a seed re-characterization? | high | answer requested; review after merge |
| #4933 | fractal#3073 | n/a | Should offline validation move into a separate declaration crate now? Is the authoring-shaped codegen model in #4300 acceptable, or must the SDK wait for a machine-readable parser grammar? Should the SDK stay in-repo first with npm later? Question 1 ownership is already settled by Johannes; #3240 is merged. | medium | answer requested |
| #4934 | fractal#3957 | n/a | Does Nathan accept a scoped deploy token from Johannes's Cloudflare account in this repository and the storybook.fractal.how domain, or which account and domain should be used? | high | answer requested |
| #4935 | fractal#3513 | n/a | May Compounding Manager assign an agent to diagnose and fix the stale-fence refusal, or does Nathan own the fix? Reply ok or mine. | medium | answer requested |
| #4936 | fractal#3623 | n/a | Which larger store write-amplification option should be measured, including the separate-SQLite-databases option, or how should the options be reframed? | high | answer requested |
| #4937 | fractal#3852 | n/a | Does Nathan accept option C in #3832: refuse all deferred message declarations from peer-bound publishers? What is the accepted residual-risk treatment for already-stored declarations whose sender binding was never recorded? Alternatively, is option B or A required? | high | answer requested; review after merge |
| #4938 | fractal#4400 | n/a | Does Nathan accept the client-v0 proposal for run and step start/completion times, remaining-duration ETA and a second arrangement for person pins? Please answer the three open questions at the end of the proposal or name changes. | medium | answer requested; review after merge |
| #4939 | fractal#3032 | n/a | Does Nathan choose A: keep st3.v1 as a frozen digest input while display and wire names change, or B: deploy a dual-digest reader before changing the writer? | high | answer requested |
| #4940 | fractal#3032 | n/a | Does Nathan want an accepting PTY reader for the renamed st3.start-operation field, and should the Fractal rename agent or Nathan implement it? | high | answer requested |
| #4941 | fractal#2964 | n/a | Does Nathan accept the all-members-upgraded publication gate for declared-priority, declared-lane and declared-budgets? Older members cannot be proven to preserve these declaration children safely. | high | answer requested |
| #4942 | fractal#2964 | n/a | Which signed stack route does Nathan accept: A, native stacks with signed local rebase and lease-protected push; B, branch-base stacks with signed merges of main; or C, sequential PRs without stacks? | high | answer requested |
| #4943 | fractal#2744 | n/a | What standing merge route does Nathan recommend for baseline rolls whose lane digests need host-level evidence? Johannes wants the simplest intended Mergemaster route, not a recurring exception. | high | answer requested |
| #4944 | fractal#2744 | n/a | Where should the host signing key live, and how should batches reach Nathan: runbook, label or fixed window? Does Nathan accept narrowing genie-freshness and ci-script-tests to true lane-definition inputs? | high | answer requested |
| #4945 | fractal#2744 | n/a | Should story registration become lane input data rather than part of the fractal-web lane definition? | high | answer requested |
| #4946 | fractal#2744 | n/a | Should scripts/install-git-hooks install into the common Git hooks directory even with core.hooksPath set, and accept FRACTAL_AGENT as appropriate? Should a contributor send the PR or should Nathan own it? | medium | answer requested |
| #4947 | fractal#2744 | n/a | Will Nathan make the person-host merge-main worker sign every pushed merge commit, or stop it from pushing merges and sign the supplied head? Does its gitconfig persist between jobs? | high | answer requested |
| #4948 | fractal#2744 | n/a | Will Nathan post a short heads-up before merging edits to required-lanes, evidence-skip-policy or lane-definition paths, so Compounding Manager can route pending evidence? | low | answer requested |
| #4949 | fractal#2744 | n/a | Should the delivery-probe and Discord-bridge self-tests join ci-script-tests or another required lane? Which lane should a contributor target? | medium | answer requested |
| #4950 | fractal#2744 | n/a | What retention should apply to message.staged, message.read, message.delivered, work.renewed and subagent.renewed, and should a contributor start the pending reader audits? | high | answer requested |
| #4951 | fractal#3476 | n/a | Does Nathan accept the attention PR1 contracts and each implementing slice after merge, including graph-designated notice runtime authority with compiled reviewed manifest hashes? Johannes has decided graph-designation; Nathan's review does not block landing through normal gates. | high | answer requested |
| #4952 | fractal#4626 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4953 | fractal#4646 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4954 | fractal#4667 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4957 | fractal#4675 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4960 | fractal#4666 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4961 | fractal#4641 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4964 | fractal#4681 / hosted #5088 | 6225232 | Does Nathan accept keeping the mocked-sleep test in the merged scheduler-independent tree timing regressions? | low | post-merge review requested |
| #4965 | fractal#4620 / hosted #4975 | 9dd9e33 | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | medium | post-merge review requested; fixes merged: #5053/#5162 and #5054/#5163 |
| #4966 | fractal#4686 | n/a | What compliant host-level evidence route does Nathan accept for this lane-definition PR? If a person-operated run is still required, will Nathan run and publish evidence for the then-current exact head? After normal gates and agent self-merge, does Nathan accept the change? Do not reuse the historical runbook head without checking it. | high | answer requested; review after merge |
| #4967 | fractal#4688 / hosted #5089 | 3455954 | Does Nathan accept this lane-definition change? flake.nix publishes packages.fractal-chatgpt with canonical bin fractal-chatgpt. packages.st-chatgpt stays the same derivation with a relative st-chatgpt symlink under live-migration ARN arn:lmig:smalltalk:2026-11-01-fractal-chatgpt-alias. scripts/chatgpt-prototype-e2e.py and the README build use fractal-chatgpt. Fractal Rename registered the alias for deletion at contraction. | low | post-merge review requested |
| #5019 | fractal#867 | n/a | Should the unattended merge service use the plugin/service-principal contract, or a narrower read-only service pairing with scoped claim-by-id, paged-claim, work-item and document reads? Nathan should name the accepted stable, revocable and rotatable non-person identity. Every write must be refused; the service must not borrow the invoking agent identity. | high | answer requested |
| #5020 | fractal#1509 | n/a | Does Nathan accept verified tar plus OCI delivery, dynamic preview allocation and bounded stateless revert as the deployment scope, with core-owned durable intent, authorization and fences and adapter-owned realization? Nathan should settle the Nix/Home Manager runtime-parameter ownership boundary, name the ARM artifact and mixed-version compatibility acceptance evidence, and identify the restart-continuity and recovery-fencing proof required before implementation. The earlier cancelled mission is not a completed continuity proof. | high | answer requested |
| #5021 | fractal#2852 | n/a | May Compounding Manager close the standalone ChatGPT prototype #2852 as superseded by merged #3050, or what unique scope should remain? The source PR stays parked until Nathan answers. | low | answer requested |
| #5022 | fractal#3032 | n/a | May Fractal adopt v0.13.0-rust.5 or a later compatible PTY release during the November 3–10 host upgrade window, and is the Fractal PTY pin freeze released for that adoption? Nathan should name the accepted version or immutable commit and any remaining host-upgrade receipt requirement. This is adoption permission, not the #209 release review or the separate start-operation accepting-reader question. | high | answer requested |
| #5039 | fractal#4872 | bc453ef | Is a code-point maxLength the right contract for the 60 existing codecs? | medium | post-merge review requested |
| #5040 | fractal#4399 | n/a | Does Nathan provide the Factory sign-off requested in the release note for the .st to .fractal configuration move, including the moved trust/configuration surface, reader list, risk classification and protected window? The change is carried by hosted copy #5018. | medium | pending merge |
| #5052 | fractal#4923 / hosted #5005 | 81d3683 | The merged #4923 stores Pins as arrangement-v2 root ordered memberships under the reserved person arrangement UUID 00000000-0000-7000-8000-000000000002 and reuses the existing place/remove/read/follow operations. #4400 had proposed v1 root placements. Is v2 membership the right storage, and is reserving ...0002 for Pins acceptable? | low | post-merge review requested |
| #5066 | fractal#4968 / hosted #5026 | a64e018 | The served KDL text is a deterministic rendering from the published revision only, not the original bytes; env values and embedded declaration KDL are redacted on the server. Is a rendering rather than the source bytes acceptable for the read-only definition view? | low | post-merge review requested |
| #5074 | fractal#4668 / hosted #5023 | f385bb6 | Is deferring only the grammar functions, while token rendering and Markdown parsing stay eager, with contained failure and retry on reopen, the right first-paint trade-off? | low | post-merge review requested |
| #5075 | fractal#4661 / hosted #5034 | bf35a7f | Does Nathan accept the tooltip accessibility fix in apps/fractal-web SidebarAgentRow.tsx:140 and the LiveHeader test wait in LiveHeader.integration.test.tsx:209? Is waiting on the served state the right test contract? | low | post-merge review requested |
| #5083 | fractal#4969 / hosted #5027 | 335bcb1 | Legacy loop shapes render in their old syntax behind the comment "// legacy loop shape; the current parser rejects it", so the served text does not round-trip through the current parser. Is labelled legacy syntax acceptable, or should those definitions render as an explicit unsupported placeholder? | low | post-merge review requested |
| #5085 | fractal#4962 / hosted #5014 | da78051 | The web Pins state follows the reserved person Pins arrangement (...0002) through arrangement-v2 memberships. A failure while the stream opens latches 'interrupted', and edits stay refused until a reopen. Is refuse-until-reopen the right edit policy, compared with retrying in the background? | low | post-merge review requested |
| #5086 | fractal#5047 / hosted #5055 | 0232bd3 | Confirm-only: Nathan signs off on the merged checker fix for the Codex finding on #4986: objective contrast markers use the visible about labels. No substantive design question is raised. | very low | post-merge review requested (confirm-only) |
| #5157 | fractal#5046 / hosted #5051 | 02729d2 | Is separating adjacent same-named queue blocks the canonical KDL rendering Nathan wants, or should the renderer merge them? | low | post-merge review requested |
| #5158 | fractal#5076 / hosted #5100 | b28ea65 | Does Nathan accept follow cancellation keyed by source, so switching the Pins source cancels the old follow? | low | post-merge review requested |
| #5159 | fractal#4644 / hosted #5080 | 6d6bbc7 | Is the Agent principal section that the schema generator now emits in docs/st3/schema.md:256-276 the intended public grammar text? Does Nathan accept fixture names config-operator/operator-two and allowlist entry operator-2_name.test? | low | post-merge review requested |
| #5160 | fractal#4648 / hosted #5081 | 7640ab4 | Does Nathan accept 386 hand-written lines in one PR instead of the proposed two-PR split: aggregation and fencing first, then terminal-first ordering and placement refinements? | medium | post-merge review requested |
| #5161 | fractal#4852 / hosted #5065 | 92b4378 | Does Nathan accept keeping the notice $defs internal-only until E3? The definitions do not add public operations or advertise runtime notice support. | low | post-merge review requested |
| #5162 | fractal#5053 | 3b402fb | Does Nathan accept this change as merged? | medium | post-merge review requested |
| #5163 | fractal#5054 | a3dde10 | Does Nathan accept this change as merged? | medium | post-merge review requested |
| #5168 | fractal#5143 | af104ce | Do these regression tests pin the right current behavior before the agent-review work starts? They cover unknown gate kinds (including agent-review) failing closed on both KDL paths and in JSON; human gate JSON bytes; human review request fields, hash and operation; and repeated request identity. | low | post-merge review requested |
| #5177 | fractal#4634 / hosted #5079 | 8286cec | Does Nathan accept this compact-width priority order: an overflow marker plus a target for each side with hidden tabs, then the add button, then the Inbox count digits, then the Inbox label text? At width 12, Inbox may show only '?2' with no label. Does Nathan also accept the size exception of 351 hand-written lines? | low | post-merge review requested |
| #5178 | fractal#4742 / hosted #5082 | 56d94de | Confirm-only: does Nathan accept the SQL ownership comment and the registry entry as merged? No substantive design question is raised. | low | post-merge review requested |
| #5179 | fractal#4858 / hosted #5070 | 9c2fbfe | Does Nathan accept the build-time schema closure as part of A1's direction? | low-to-medium | post-merge review requested |
| #5199 | fractal#4631 / hosted #5136 | ae0384e | Does Nathan accept the Usage pane reading served budgets and objectives only, with no client-side budget math? | medium-low | post-merge review requested |
| #5200 | fractal#5029 / hosted #5041 | c7fdda5 | Is the redaction set, with credentials and env values removed before canonical KDL is served, the client-v0 contract Nathan wants? | medium | post-merge review requested |
| #5201 | fractal#4651 / hosted #5024 | 1c2e3e2 | Does Nathan accept sidebar answers through the existing ask-answer mutation without a confirmation step? | low-medium | post-merge review requested |
| #5202 | fractal#5092 / hosted #5123 | 404b20b | Does Nathan accept a single retry per chunk, bound to the build, then an honest error state with no auto-reload? | low | post-merge review requested |
| #5203 | fractal#5140 / hosted #5152 | f41267a | Does Nathan accept prefetch at idle and on hover or focus, with Recent changes opening in a fixed-height box? | low | post-merge review requested |
| #5210 | mergemaster#58 | 01a87dea048a14f1ceac378dd26c8021ab05bf80 | Is binding the merge-group candidate by run head_sha (Git-resolved tree) plus an exact gh-readonly-queue//pr-- name sufficient? The queue-name hash is not compared with the candidate, and the merge_group base_sha ancestry check lives in the gate, not in this verifier. | low today | post-merge review requested |
| #5223 | fractal#5093 / hosted #5124 | 46c5fb8 | Is a three-engine browser proof of the lazy retry the right permanent CI cost? The tests cover Chromium, Firefox and WebKit. | low | post-merge review requested |
| #5224 | fractal#5094 / hosted #5125 | 982a0d2 | Does Nathan accept one retry, then a reload prompt, for conversation and terminal panes? | low | post-merge review requested |
| #5225 | fractal#5164 / hosted #5176 | cc692e4 | Is a fingerprint-recognized LICENSE file acceptable license evidence when the manifest has no license field? The file must match a recognized license byte for byte. Explicit invalid metadata still fails. | low-medium | post-merge review requested |
| #5226 | fractal#5169 / hosted #5180 | 0d1ad7f | Does Nathan accept English-only React Aria strings until real internationalization? Number grouping and percentages follow en-US even in non-English browsers. | low | post-merge review requested |
| #5244 | fractal#4859 / hosted #5220 | ca4b86a | The Rust validator accepts leap second 23:59:60, but the Effect codec rejects every :60. Which convention should both follow? | medium | post-merge review requested |
| #5245 | fractal#5172 / hosted #5193 | 18f5130 | Confirm-only: does Nathan accept the merged SQL person-hold, nearest-rank quartile and keyed-trigger EXPLAIN test coverage for the ETA projection? The ETA design questions stay in #4938. | low | post-merge review requested |
| #5246 | fractal#5198 / hosted #5204 | b148972 | Is it acceptable that the metadata cache falls back to regeneration only for four English git apply.c diagnostics: 'patch failed: :', ': patch does not apply', ': does not exist in index', and ': already exists in index', run with LC_ALL=C and LANGUAGE=C? Every other git output fails the tick loudly, including a stale binary hunk or mode change, as accepted by Fractal Product Lead. Does Nathan instead want a structured signal, such as git apply --reject or patch preimage-blob comparison? | low | post-merge review requested |
| #5247 | fractal#5131 | b79237f | Does Nathan accept these surfaces as merged: a hidden read-only 'fractal migrate launchd-domain' command that Home Manager uses to discover the domain, and a Home Manager-managed st3/launchd-staged marker only during the macOS stage phase? They let the guard work without a daemon restart at the stage-to-completed switch. Removing the marker prints one harmless 'rmdir ... Not a directory' line at the first completed activation. A completed host configured for the launchd user domain without a journal now refuses activation rather than falling back to the GUI domain. Does Nathan want the command documented or public, the marker elsewhere, or a supported fresh user-domain install path instead? | medium | post-merge review requested |
| #5248 | fractal#5153 / hosted #5214 | d10b0cf | Does Nathan accept the web acquiring @fractal/sdk ./pty and ./mutations lazily on first use for terminal attach, send and folder edits, while one shared read runtime stays eager? This includes the three-line Rolldown $initial Effect group that keeps eager Effect code in one chunk so new lazy roots do not fragment it. | medium-low | post-merge review requested |
| #5249 | fractal#5216 | 04c0eeb | Confirm-only: does Nathan accept the generated v0.3.26 compatibility baseline output as merged? The release workflow produced the baseline. Compounding Manager opened the PR by hand only because Actions PR creation failed with HTTP 403 (#3871). The separate PR-opening policy question belongs to #3871, not this baseline review. | low | post-merge review requested |
| #5261 | fractal#5233 / hosted #5241 | 1691205 | Confirm-only: Nathan signs off on the merged Swift fixes from #5233, including #5229. No substantive design question is raised. | medium | post-merge review requested |
| #5262 | fractal#5217 / hosted #5222 | 625f154 | Is runtime.refresh as a durable claim kind, rather than local-only state, the representation Nathan wants for refresh episodes, given the replication cost? | low | post-merge review requested |
| #5263 | fractal#5038 / hosted #5101 | 4987bc3 | Does Nathan accept a Pins section ordered by the served order and shown above folders, with pin actions in the row menu? Pins defaults remain an open product question with Johannes. | low-medium | post-merge review requested |
| #5264 | fractal#5042 / hosted #5149 | 555803e | Does Nathan accept the declaration dialog showing the canonical KDL as served, with no client-side formatting? | low | post-merge review requested |
| #5267 | fractal#5256 | 56b22d4 | Should the agent-review evidence types live in the fractal-kernel model, as they do now in the private child module model/evidence_contract.rs, or in a separate review module? | low | post-merge review requested |
| #5274 | fractal#5120 / hosted #5195 | c57fac2 | Is it acceptable that eta samples are shared across missions by (step definition hash, step path), so that identical steps in different missions pool their duration history, instead of being scoped per mission? | medium | post-merge review requested |
| #5275 | fractal#4692 / hosted #5097 | bcc453f | Does Nathan accept the foreground phone-command contract (phone-commands v1; inspect, navigate, message.read with message_id; member-local in-memory queue max 8/device; default 30s/max 60s TTL adjudicated by daemon; held GET result), or what change is required? | medium | post-merge review requested |
| #5276 | fractal#4864 / hosted #5012 | 0b81147 | Do you want such cohort alignments (Effect minor jump plus regeneration) as their own PR class with a mandatory maintainer pre-review, or is retroactive review enough? | medium | post-merge review requested |
| #5282 | fractal#5271 | 3f552a6 | Is #[doc(hidden)] pub on the kernel evidence_contract module the right cross-crate exposure, or does Nathan prefer hidden re-exports of the three types? | low | post-merge review requested |
| #5291 | mergemaster#59 | 999d784476408368bdf08078542021c14f2df224 | Does Nathan accept the merged caller-bound merge_group shadow gate? The gate binds the event repository, head_sha, base_sha and the gh-readonly-queue head_ref to base_ref. It marks every Decision expectation_source=caller. It reports a would-satisfy result as refuse/report_only_queue_parent, because the event base_sha is the queue parent, not the live accepted BASE. If accepted: should a later enforcing mode obtain the live BASE from the protected branch tip at decision time, or from the Namespace signer-input record? | low | post-merge review requested |
| #5292 | fractal#5268 / hosted #5269 | 3ff707a | Is a per-driver synchronous admission gate the right layer for refresh fencing, versus the harness-state transition counters? | low | post-merge review requested |
| #5293 | fractal#5213 / hosted #5257 | 21f7f9f | Is gating claim admission on a per-agent seal in the writer transaction acceptable, or does Nathan prefer the check in the reconcile layer? | medium | post-merge review requested |
| #5294 | fractal#5138 / hosted #5270 | 95c5444 | Does Nathan accept a sticky plain-text fallback for the session after a highlighter chunk failure, instead of a retry? | low | post-merge review requested |
| #5301 | fractal#5284 | 5634120 | Does Nathan accept this alignment with the server cursor-only continuation contract, or is an iOS pagination change required? | low | post-merge review requested |
| #5306 | fractal#5287 | 806ad12 | Should a retired reviewer stay terminal even when a later publication redeclares the same agent (current behavior; recovery is a revision with a different reviewer)? | medium | post-merge review requested |
| #5307 | fractal#5278 / hosted #5290 | 3387806 | 1. Native user commands (! shell, Python): the producer has no completion event, so one allowed native command makes that incarnation refresh-unsupported until it is replaced. Is that acceptable for managed agents, or should the gate treat a returned user-command result message as completion? 2. Settlement: a dispatch settles only on its own tool_result, an explicit approval denial, or a matching terminal toolResult message. Overlapping calls with the same name stay busy until each settles. Do you see a producer path where an admitted call gets none of these (that path would block refresh permanently)? 3. Boundary: text-only native turns may start after close because they cannot run tools. Do you agree that turn-start admission is not needed for safety, given the drafted upstream request for fail-closed turn admission? |
low live / medium design | post-merge review requested |
| #5309 | fractal#5288 / hosted #5296 | 4975dd5 | Do you accept the hand-built tabpanel (the PhoneWorkspaceTabs pattern) instead of RAC TabPanel, as the way to avoid useHasTabbableChild's DOM scans? | low-medium | post-merge review requested |
| #5314 | fractal#4693 / hosted #5286 | 6cdf25d | Does Nathan accept these typed phone-command client APIs and held-GET deadline policy, or what client compatibility change is required? | medium | post-merge review requested |
| #5315 | fractal#5289 / hosted #5297 | f2fed6c | Do you accept declaring exact heights for fixed row kinds (from the shared CSS geometry) in the virtualizer, instead of measuring every row? | low-medium | post-merge review requested |
| #5316 | fractal#5243 / hosted #5255 | 08ea928 | Confirm-only: Nathan signs off on the merged Swift client CI lane. No substantive design question is raised. | low | post-merge review requested |
| #5328 | fractal#5321 | 32f7e69 | Does Nathan accept the command lifecycle, collections-only foreground classification and terminal-retention/idempotency policy, or what change is required before later wiring activates it? | medium | post-merge review requested |
| #5329 | fractal#5305 / hosted #5320 | e6c82b4 | Is it right that the projection follows pty-core's identity policy (Match and Unknown keep the PID) at this fence, or should Fractal keep its own stricter policy? | medium | post-merge review requested |
| #5330 | fractal#5300 / hosted #5308 | 37a64fb | OMP cannot capture or rebind a pending native prompt, so a pending prompt keeps the agent busy, and a prompt-restoration timeout makes it permanently unsupported until replacement (sticky lostPrompt). Is a permanent block right for a restoration timeout, or should a later normal resolution of the same session/call/tool key clear it? | low | post-merge review requested |
| #5331 | fractal#5272/#5279 / hosted #5280/#5281 | 0c779bc; a468b96; c5c98f5 | Does Nathan accept the bounded terminal projection that replaced candidates_sql in the #5299 resolution? | medium | post-merge review requested; rollback-trigger follow-up #5341 merged |
| #5332 | mergemaster#60 | 423c28aa667c01159a3a15a854760e194b891c1c | Confirm-only: Nathan signs off on the merged advisory tests. No substantive design question is raised. | low | post-merge review requested |
| #5337 | fractal#5324 | 4dc11b5 | Is 'any agent that wrote a work.* event on the step may not review it' the right independence rule? It excludes more agents than necessary, on purpose. | medium | post-merge review requested |
| #5338 | fractal#5128 / hosted #5327 | f811eb1 | Confirm-only: Nathan signs off on the merged integrator capacity precheck. No substantive design question is raised. | low-medium | post-merge review requested |
| #5355 | fractal#5343 | f522474 | Should a reviewer be excluded for work.* events on ANY earlier step in the carry chain (current behavior; it excludes more agents than necessary, on purpose), or only on the directly carried source step? | medium | post-merge review requested |
| #5356 | fractal#5350 | 275b5f8 | Confirm-only: Nathan signs off on the merged delegation regression tests for agent-review slice 5c. No substantive design question is raised. | low | post-merge review requested |
| #5357 | mergemaster#61 | e8edf0e324d2074c8a418aeeb291462070c7f16f | Does Nathan accept the merged gate --event queue behavior? Nathan should focus on gate.rs, the queue branch and human_approval_only. | low-medium | post-merge review requested |
| #5358 | fractal#5319 / hosted #5325 | c13d49d | Confirm-only: Nathan signs off on the generated public Swift NoticeModels API and decode validation. The leap-second convention is the same as Rust; the cross-language convention question remains in #5244. | medium-low | post-merge review requested |
| #5359 | fractal#5230 / hosted #5231 | a996aa9; 8a3bc64 | Does Nathan accept the export name @fractal/client/schema/read and the new failure mode for malformed terminal frames? | medium | post-merge review requested |
| #5368 | fractal#5228 / hosted #5265 | 1fec09b | Is batched adoption (older rows can appear over about 8 frames) acceptable UX, compared to one synchronous long frame? | medium | post-merge review requested |
| #5369 | fractal#5146 / hosted #5349 | 055d862 | Do you accept 'Finishing' as the label when the server reports zero remaining-step samples, instead of hiding the ETA? | low | post-merge review requested |
| #5370 | fractal#5317 / hosted #5342 | 864c86c | Is the reinstall after a directory loss (dev+inode identity plus DELETE_SELF/IGNORED) enough on macOS FSEvents, where the event semantics differ from inotify? | medium | post-merge review requested |
Inventory dispositions
- pty#214: Closed. Nathan recorded the decision: off-terminal attach restarts only for y/yes; EOF declines. feat(ding): show sender relationship in notices #215 merged. No unanswered sign-off request remains.
- pty#204: Closed as answered. pty runs once by default; reported detach respawn belongs to OMP. list_dir still drops errors yielded during iteration: entries.flatten() hides a partial inbox listing #214/feat(ding): show sender relationship in notices #215 and DING relationship marker can be confident from an incomplete observation: completeness is checked in a second catalog traversal #216 address prompt edges.
- fractal#4713 and Install ownership receipts for script and setup installers (A1) #4762: Their hosted changes are represented once by fix(web): provide Python for the web lane regression suite (hosted #4713) #4847 and Install ownership receipts for script and setup installers (hosted #4762) #4848; no duplicate child.
- fractal#4841 and Advance compatibility baselines to v0.3.24 #4740: Manual baseline PRs are evidence for release-daily: advance-compatibility-baselines cannot open its PR (Actions may not create PRs) #3871, not new questions.
- fractal#3240: Merged; remaining declaration questions are tracked against Graph as code: offline
st validateand a generated TypeScript declaration SDK #3073, with Add @fractal/declarations: generated TypeScript declaration SDK #4300 as evidence. - fractal#2788: Merged; old close request is obsolete.
- fractal#937: Closed; replacement/dependency question is retained under Add ordered terminal input sessions with durable owner audit (D18) #1335.
- fractal#3458: Same assistant/mcp schema and rollout question is tracked once under feat(st3): st-mcp stdio MCP server with its own assistant principal #3531.
- fractal#3832: Same deferred-sender boundary is tracked once under fix(api): bind message senders to the peer-bound agent #3852.
- fractal#3837: Factory explicitly delegated repair to Fractal Web 2 and requested Factory review; no unanswered Nathan decision in its latest comments. Merge queue: per-entry change detection skips kit runs for earlier batch entries #3457 retains the explicit CI-owner questions.
- fabric#340, Resync events: notify a live agent when a declared resource carrier changes on disk #341 and feat(resync): notify a live agent when a declared resource carrier changes #345; fractal#3644: Fabric PRs merged and v0.9.4 exists. The review/tag prerequisite request is completed; Clarify Fabric rekey boundary and compare identity cutover designs #3644 is superseded by the rekey stack in the current watch conditions.
- fractal#3514: Merged; deployment permission question remains under Hosted Storybook: enable the existing deploy at storybook.fractal.how #3957.
- fractal#3608: Merged; the old baseline-CPU request is not carried as a new open sign-off.
- fractal#1556, Warn before apply restarts and defer launch changes to safe points #1639, feat(st3): migrate arrangement placements without replacing winners #1995, Fractal: agent messages and folded outgoing delivery rows #2333, fractal: native history paging with a virtualized transcript #3098, perf(st-runtime): share one PTY registry metadata cache across snapshots #3489 and Spike: drive Fractal-managed Codex agents through t3code ACP #4143: Latest digest assigns readiness, repair or coordinator work, not an additional explicit Nathan question. Normal review gates still apply.
- fractal#1477 and Expose authorized native subjects with registry-derived client contracts #1464 newer Intake findings: Retained in their existing contract children, not duplicated as new items.
- Private security tracker referenced by the digest: Excluded from this public tracker to preserve confidentiality. Its current private disposition was NOT RUN; no private details were fetched or published.
- Expired watches evidence: one ref layout for st evidence put and script producers; head/host/signer-role bindings in evidence-verify output #3529, evidence design question: protected producers (unforgeable by same-uid agents) and sticky NO-GO SOURCE snapshots #3530, Categorized diff stats: st-owned classification for Captured / Workspace / PR counts #3542 and Contract intake: harness child (sub-agent) sessions as addressable conversations #2288: Watch expiry alone is not a question to Nathan.
- Older superseded close proposals listed as removed in the Q53 digest: Already removed by the source digest; no new sign-off requested.
- PTY restart-environment adoption request: The initial exclusion was too broad. Release review does not settle upstream Fractal adoption or its bundled-pin freeze. The distinct adoption question is now tracked in Sign-off: fractal#3032 PTY adoption and pin-freeze release #5022. Current dotfiles re-pin receipts approve their named pin change but explicitly leave the bundled internal PTY unchanged and grant no rollout authority.
- Person-route batch 2 fix(web): provide Python for the web lane regression suite #4713 and Install ownership receipts for script and setup installers (A1) #4762: Replaced by merged hosted copies fix(web): provide Python for the web lane regression suite (hosted #4713) #4847/Install ownership receipts for script and setup installers (hosted #4762) #4848; Persist Home Manager install ownership receipts (A2) #4620 remains a separate open child.
Evidence scope
Compounding Manager read the latest curator digest and question comments in #2744, current watch conditions, and the latest available upstream sweep reports dated 2026-10-22 and 2026-10-21. Older local reports supplied leads only; current GitHub state determined inclusion. The current watch conditions also supplied the attention slice post-merge request and the stale-fence ownership question. GitHub API data supplied merge SHAs and confirmed that v0.13.0-rust.6 is absent and Fabric v0.9.4 exists.
NOT RUN: code tests, builds, CI reruns or dispatches, merges, release tags, permission changes, private security-item inspection, and acceptance testing of the underlying changes. Creating this tracker does not assert that any open PR currently passes its gates.
Additional questions from the 2026-11-04 sweep
Compounding Manager rechecked every existing sub-issue before adding these questions. Each checkbox remains open until Nathan answers or signs off.
- Sign-off: fractal#867 read-only merge-service identity #5019 — read-only merge-service identity (high).
- Sign-off: fractal#1509 deployment scope and ownership #5020 — deployment scope and ownership (high).
- Sign-off: fractal#2852 superseded ChatGPT prototype disposition #5021 — superseded ChatGPT prototype disposition (low).
- Sign-off: fractal#3032 PTY adoption and pin-freeze release #5022 — PTY adoption and pin-freeze release (high).
Attention and configuration follow-up
- Sign-off: fractal#4872 code-point maxLength contract #5039 — Nathan answers the code-point maxLength question for the 60 existing codecs; Preserve JSON Schema string semantics in Effect codecs #4872 is merged.
- Sign-off: fractal#4399 Factory sign-off for .fractal config move #5040 — Nathan provides the requested Factory sign-off for the configuration move; hosted Move repository configuration from .st to .fractal (hosted #4399) #5018 is pending merge.
Compounding Manager updated #4904 to post-merge review for hosted #4981 at ea2ce38c71a465ba2e4d4719ddf9488cb48ea8be on 2026-11-04. Its question is unchanged.
Install receipt follow-ups
The existing sign-off sub-issues #4901 (#4848 / #4762) and #4965 (#4620 / hosted #4975) record both valid P2 receipt findings. Both are medium risk. Fractal Rename merged fixes #5053 and #5054; their post-merge reviews are #5162 and #5163. Nathan can review the linked fixes with the original items. Fractal Rename owns follow-up answers. Merge does not close any sign-off item or assert release delivery.
Merged-change review follow-up
- Sign-off: fractal#5228 batched older-message adoption #5368 — web: load earlier messages from served history #5228 / hosted web: load earlier messages from served history (hosted #5228) #5265: batched older-message adoption (medium).
- Sign-off: fractal#5146 served mission ETA Finishing label #5369 — web: show served mission ETA on mission rows #5146 / hosted web: served mission ETA (hosted #5146) #5349: served mission ETA Finishing label (low).
- Sign-off: fractal#5317 R2 shared file-subscription reinstall policy #5370 — fix(reconcile): share reference-counted file subscriptions #5317 / hosted fix(reconcile): share reference-counted file subscriptions (hosted #5317) #5342: R2 shared file-subscription reinstall policy (medium).
- Sign-off: fractal#5343 agent-review slice 5b carry-chain independence #5355 — fractal#5343: agent-review slice 5b carry-chain independence (medium).
- Sign-off: fractal#5350 agent-review slice 5c delegation regression #5356 — fractal#5350: agent-review slice 5c delegation regression (low).
- Sign-off: mergemaster#61 caller-bound report-only queue event gate #5357 — mergemaster#61: caller-bound report-only queue event gate (low-medium).
- Sign-off: fractal#5319 A3 generated Swift notice models #5358 — fractal#5319: A3 generated Swift notice models (medium-low).
- Sign-off: fractal#5230 read-only schema export and terminal decode failure mode #5359 — fractal#5230: read-only schema export and terminal decode failure mode (medium).
- Sign-off: fractal#5324 agent-review slice 5a independence rule #5337 — feat(store): add internal agent-review step responsibility snapshot #5324: agent-review slice 5a independence rule (medium).
- Sign-off: fractal#5128 integrator capacity precheck confirmation #5338 — fix(web): precheck integrator capacity before candidate writes #5128: integrator capacity precheck confirmation (low-medium).
- Sign-off: fractal#5321 dormant phone command lifecycle #5328 — fractal#5321: dormant phone command lifecycle (medium).
- Sign-off: fractal#5305 incremental PTY registry projection identity policy #5329 — fractal#5305: incremental PTY registry projection identity policy (medium).
- Sign-off: fractal#5300 R2 OMP pending ask preservation #5330 — fractal#5300: R2 OMP pending ask preservation (low).
- Sign-off: fractal#5272 bounded terminal ETA projection resolution #5331 — fractal#5272/fix(store): claim-free ETA publication generation and rollback-safe terminal index #5279 / hosted fix(store): index-bounded ETA candidates and cross-mission card invalidation (hosted #5272) #5280/fix(store): claim-free ETA publication generation and rollback-safe terminal index (hosted #5279) #5281: bounded terminal ETA projection resolution (medium; pair and low-risk rollback-trigger follow-up fix(store): keep the v4 ETA trigger layout across a fresh-store rollback (hosted #5340) #5341 merged).
- Sign-off: mergemaster#60 advisory tests confirmation #5332 — mergemaster#60: advisory tests confirmation (low).
- Sign-off: fractal#4693 typed phone-command clients and held GET deadline #5314 — Phone commands: client methods and prose contract #4693 / hosted Phone commands: client methods and prose contract (hosted #4693) #5286: typed phone-command clients and held GET deadline (medium).
- Sign-off: fractal#5289 Q87 exact fixed-row virtualizer heights #5315 — perf(web): skip initial measurement of exact RM1 agent rows #5289 / hosted perf(web): skip initial measurement of exact RM1 agent rows (hosted #5289) #5297: Q87 exact fixed-row virtualizer heights (low-medium).
- Sign-off: fractal#5243 non-required Swift client CI lane #5316 — ci(swift): add a hosted Linux package-test lane #5243 / hosted ci(swift): add a hosted Linux package-test lane (hosted #5243) #5255: non-required Swift client CI lane (low).
- Sign-off: fractal#5288 Q87 sidebar tabpanel implementation #5309 — perf(ui): avoid sidebar tab-panel visibility scans #5288 / hosted perf(ui): avoid sidebar tab-panel visibility scans (hosted #5288) #5296: Q87 sidebar tabpanel (low-medium).
- Sign-off: fractal#5287 agent-review slice 4 retired reviewer policy #5306 — feat(store): add internal agent-review reviewer resolver #5287: agent-review slice 4 retired reviewer policy (medium).
- Sign-off: fractal#5278 R2 OMP adapter refresh gate #5307 — drivers: gate managed OMP tool effects during idle refresh #5278: R2 OMP adapter refresh gate (low live / medium design).
- Sign-off: fractal#5284 iOS cursor-only timeline continuation #5301 — fix(ios): send only the cursor on timeline continuations #5284: iOS cursor-only timeline continuation (low).
- Sign-off: mergemaster#59 caller-bound merge_group shadow gate #5291 — mergemaster#59: caller-bound merge_group shadow gate (low).
- Sign-off: fractal#5268 R2 per-driver refresh admission gate #5292 — fractal#5268: R2 per-driver refresh admission gate (low).
- Sign-off: fractal#5213 R2 writer-transaction intake seal #5293 — fractal#5213: R2 writer-transaction intake seal (medium).
- Sign-off: fractal#5138 highlighter chunk session fallback #5294 — fractal#5138: highlighter chunk session fallback (low).
- Sign-off: fractal#5271 agent-review grammar helper exposure #5282 — feat(missions): add internal agent-review grammar helper #5271: agent-review slice 3 grammar helper exposure (low).
- Sign-off: fractal#5120 ETA shared cross-mission duration samples #5274 — Serve run ETA on client-v0 mission summaries #5120 / hosted Serve run ETA on client-v0 mission summaries (hosted #5120) #5195: ETA shared cross-mission duration samples (medium).
- Sign-off: fractal#4692 foreground phone-command contract #5275 — Phone commands: contract and generated model surfaces #4692 / hosted Phone commands: contract and generated model surfaces (hosted #4692) #5097: foreground phone-command contract (medium).
- Sign-off: fractal#4864 effect-utils and Effect dependency alignment #5276 — chore(deps): align effect-utils to published f77b58a4 #4864 / hosted chore(deps): align effect-utils to published f77b58a4 (hosted #4864, v2) #5012: effect-utils and Effect dependency alignment (medium).
- Sign-off: fractal#5256 agent-review evidence contract module #5267 — feat(kernel): add internal agent-review evidence contract #5256: agent-review slice 2 evidence-contract module (low).
- Sign-off: fractal#5233 Swift fixes including #5229 #5261 — fix(swift): preserve optional agent relationship data #5233 / hosted fix(swift): preserve optional agent relationship data (hosted #5233) #5241: Swift fixes including test(swift): make fixture checks portable and order independent #5229 (medium).
- Sign-off: fractal#5217 durable runtime.refresh representation #5262 — Persist refresh episodes with deterministic age reduction #5217 / hosted Persist refresh episodes with deterministic age reduction (hosted #5217) #5222: durable runtime.refresh representation (low).
- Sign-off: fractal#5038 Pins sidebar section and row actions #5263 — web: render pinned sidebar section and pin actions #5038 / hosted web: render pinned sidebar section and pin actions (hosted #5038) #5101: Pins sidebar section and row actions (low-medium).
- Sign-off: fractal#5042 served canonical mission KDL dialog #5264 — web: display served canonical mission KDL #5042 / hosted web: display served canonical mission KDL (hosted #5042) #5149: served canonical mission KDL dialog (low).
- Sign-off: fractal#5216 v0.3.26 compatibility baseline confirmation #5249 — Roll compatibility baselines to v0.3.26 #5216: v0.3.26 compatibility baseline confirmation (low).
- Sign-off: fractal#5153 lazy SDK services and initial Effect grouping #5248 — web: load SDK terminal and mutation services on first use #5153 / hosted web: load SDK terminal and mutation services on first use (hosted #5153) #5214: lazy SDK services and initial Effect grouping (medium-low).
- Sign-off: fractal#5131 launchd staging and Home Manager guard surfaces #5247 — fix(migrate): use the discovered launchd domain for staging and Home Manager guards #5131: launchd staging and Home Manager guard surfaces (medium).
- Sign-off: fractal#5198 metadata-cache diagnostic fallback policy #5246 — fix(web): fail on corrupt integrator metadata cache entries instead of regenerating #5198 / hosted fix(web): fail on corrupt integrator metadata cache entries instead of regenerating (hosted #5198) #5204: metadata-cache diagnostic fallback policy (low).
- Sign-off: fractal#5172 ETA SQL regression coverage confirmation #5245 — test(store): cover run ETA person holds and quartiles through SQL #5172 / hosted test(store): cover run ETA person holds and quartiles through SQL (hosted #5172) #5193: ETA SQL regression coverage confirmation (low).
- Sign-off: fractal#4859 A2b Rust notice timestamp convention #5244 — codegen: emit internal typed Rust notice models (A2b) #4859 / hosted codegen: emit internal typed Rust notice models (A2b) (hosted #4859) #5220: A2b Rust notice timestamp convention (medium).
- Sign-off: fractal#5093 three-engine lazy retry proof #5223 — web: prove lazy retry across Chromium Firefox and WebKit #5093 / hosted web: prove lazy retry across Chromium Firefox and WebKit (hosted #5093) #5124: three-engine lazy retry proof (low).
- Sign-off: fractal#5094 conversation and terminal chunk recovery #5224 — web: recover conversation and terminal chunk failures honestly #5094 / hosted web: recover conversation and terminal chunk failures honestly (hosted #5094) #5125: conversation and terminal chunk recovery (low).
- Sign-off: fractal#5164 recognized LICENSE fingerprint evidence #5225 — ci(web): accept a verified LICENSE file when a manifest has no license field #5164 / hosted ci(web): accept a verified LICENSE file when a manifest has no license field (hosted #5164) #5176: recognized LICENSE fingerprint evidence (low-medium).
- Sign-off: fractal#5169 English-only React Aria strings #5226 — web: ship only English React Aria strings #5169 / hosted #5180: English-only React Aria strings (low).
- Sign-off: mergemaster#58 namespace signer verification candidate binding #5210 — mergemaster#58: namespace signer verification candidate binding (low today; signer activation prerequisites remain).
- Sign-off: fractal#4631 served usage budgets, objectives and history #5199 — #4631 / hosted #5136: served usage budgets, objectives and history (medium-low).
- Sign-off: fractal#5029 canonical redacted mission declarations #5200 — #5029 / hosted #5041: canonical redacted mission declarations (medium).
- Sign-off: fractal#4651 sidebar alert answers #5201 — #4651 / hosted #5024: sidebar alert answers (low-medium).
- Sign-off: fractal#5092 build-bound lazy chunk recovery #5202 — #5092 / hosted #5123: build-bound lazy chunk recovery (low).
- Sign-off: fractal#5140 lazy Recent changes registry #5203 — #5140 / hosted #5152: lazy Recent changes registry (low).
- Sign-off: fractal#4634 compact TUI tab-strip priorities #5177 — #4634 / hosted #5079: compact TUI tab-strip priorities (low).
- Sign-off: fractal#4742 replay dispatch and handoff readers confirmation #5178 — #4742 / hosted #5082: replay dispatch and handoff readers confirmation (low).
- Sign-off: fractal#4858 A2a build-time notice schema closure #5179 — #4858 / hosted #5070: A2a build-time notice schema closure (low-to-medium).
- Sign-off: fractal#5143 agent-review slice 1 regression contracts #5168 — #5143: agent-review slice 1 regression contracts; Q17 decisions recorded (low).
- Sign-off: fractal#4688 fractal-chatgpt package and migration alias #4967 — #4688 / hosted #5089: fractal-chatgpt package and migration alias (low).
- Sign-off: fractal#5053 retained macOS installer receipt repair #5162 — fix(install): refresh ownership after retained macOS installers (hosted) #5053: retained macOS installer receipt repair (medium).
- Sign-off: fractal#5054 proven Home Manager rollback receipt repair #5163 — fix(hm): accept a stale receipt only on proven Home Manager rollback (hosted) #5054: proven Home Manager rollback receipt repair (medium).
- Sign-off: fractal#4681 scheduler-independent test timing #4964 — #4681 / hosted #5088: keep mocked-sleep timing test (low).
- Sign-off: fractal#5046 canonical KDL adjacent queue blocks #5157 — #5046 / hosted #5051: canonical KDL adjacent queue blocks (low).
- Sign-off: fractal#5076 Pins follow source cancellation #5158 — #5076 / hosted #5100: Pins follow source cancellation (low).
- Sign-off: fractal#4644 Agent principal grammar #5159 — #4644 / hosted #5080: Agent principal grammar (low).
- Sign-off: fractal#4648 exec-only state aggregation and fencing size #5160 — #4648 / hosted #5081: exec-only state aggregation and fencing size (medium).
- Sign-off: fractal#4852 internal-only notice schema definitions #5161 — #4852 / hosted #5065: internal-only notice schema definitions (low).
Posted on behalf of @schickling
| field | value |
|---|---|
agent_identity |
unknown |
session |
unknown |
agent_persona |
unknown |
agent_supervisor |
unavailable |
agent_tool |
unknown |
agent_tool_version |
unknown |
agent_runtime |
unknown |
agent_model |
unknown |
worktree |
unknown |
Metadata
Metadata
Assignees
Labels
No labels
Activity
74 remaining items
schickling-assistant commentedon Nov 7, 2026
@myobie, a question from Compounding Manager (
agent/compoundingtech/compounding-manager) about macOS CI:.github/workflows/macos.yml(workflow 371644653) has beendisabled_manuallysince 2026-10-13. We found no recorded reason. Who disabled it, and why: cost, flakiness or retirement?An upcoming daemon change has macOS-only code paths. Until we have your answer, its macOS tests run on a fleet Mac, and the full output is posted on the PR. Would you accept re-enabling
macos.ymlfor one focused, path-filtered job (cargo test of the touched crates only, about 10-15 min on macos-14)? If you prefer that macOS CI stays off, the recorded fleet-Mac run stays the route.Posted on behalf of @schickling
agent_identitysessionagent_personaagent_supervisoragent_toolagent_tool_versionagent_runtimeagent_modelworktreeschickling-assistant commentedon Nov 7, 2026
@myobie, three product questions from Compounding Manager (
agent/compoundingtech/compounding-manager). The merged PRs are linked for your post-merge review:Posted on behalf of @schickling
agent_identitysessionagent_personaagent_supervisoragent_toolagent_tool_versionagent_runtimeagent_modelworktreeschickling-assistant commentedon Nov 8, 2026
Problem
The wind-down inventory retained three possible kit policy gaps from #5147: misspelled reserved tags silently falling back, deletion of a tagged story silently reducing coverage, and five unexplained legacy exclusions. Reopening those as new work would duplicate the later fix.
Motivation
Preserve intentional CI coverage and discoverable review context. agent/dotfiles/fractal-web-2's retirement must neither lose the three questions nor restart changes that already landed. #4898 requires Nathan's review to remain distinct from implementation completion.
Current state on main
Fetched source snapshot:
f5fb43f9cd7c20bd2726e42cf8ae8a9fd6e2bdc3, 2026-11-08. #5665 merged on 2026-11-07 at 18:07:44Z; #5147 is also MERGED.packages/fractal-ui/scripts/kit-play-index.mjs:4,35-44freezes the vocabulary (merge-only,both-widths) and rejects unknown reservedmerge-/both-tags before lane selection. Unrelated tags remain allowed.packages/fractal-ui/src/scenario.story-check.test.ts:222-244composes the configured source stories, compares exact expected IDs, names missing IDs and rejects replacement/shrunk sets. This is an identity-level guard, not a story-count assertion.packages/fractal-ui/.storybook/stories.ts:4documents why.legacy-stories.tsxfiles do not match the shipping.stories.tsxglob. ci(web): guard kit policy vocabulary and lane membership #5665 identifies the five older CSFs renamed by ci(web): derive kit play sets and story paths from tags (hosted #4752) #5147.Contract and constraints to keep
Reserved-tag typos must fail closed, intended lane membership changes must be explicit and reviewable, and legacy exclusions must remain documented. A same-count replacement must not hide lost coverage. Do not weaken a benchmark or turn an intentional story removal into a silent baseline regeneration.
Acceptance
For wind-down disposition, retain these three source receipts under #5147/#5665 and mark K03 as implemented on inspected main, subject to the existing review process. No new implementation issue is justified by the inventory alone.
When coverage changes intentionally, the normal reviewed diff must name changed IDs and update the reviewed membership list. Reopen only a concrete regression demonstrated against these guards; do not close Nathan's review merely because #5665 merged.
Evidence and history links
Open questions
Does Nathan accept the reserved-tag policy, exact membership-change signal and deliberate legacy exclusions? Keep that review question in #4898; the main-source recheck did not identify a remaining implementation gap.
Posted on behalf of @schickling
agent_identitysessionagent_personaagent_supervisoragent_toolagent_tool_versionagent_runtimeagent_modelworktree