<!-- CURSOR_AGENT_PR_BODY_BEGIN -->
## Metric
This PR moves the worker-direction metric: whether autonomous agents
read a written, dated current priority before picking work.
| | Before | After |
| --- | --- | --- |
| Written current priority in files workers read | 0 (two #303 comments
on 2026-10-13 asked for three items; workers shipped #328–#334 instead)
| 1 dated section in `AGENTS.md` and `docs/north-star.md`, plus a
pointer in `docs/north-star-progress.md` |
It does not change `supervisor.lock` flake rate, #280, or the
two-machine journey. Those remain the only in-scope product work.
## What changed
Docs and agent instructions only. No source or test changes. Open PRs
#303, #280, and #334 are untouched.
Inspected for task selection: `AGENTS.md` (always applied), `CLAUDE.md`
(pointer to `AGENTS.md`), `.algal/` (product manifests, not agent
config), `docs/north-star.md`, `docs/north-star-progress.md`,
`docs/main-policy.md` (PR enforcement, not task pick), and repo
skills/routines (none pick the next task).
Added a short **Current priority (set 2026-10-15)** section:
1. Root-cause and fix the intermittent `supervisor.lock` refusal on
Linux (#303), with a same-inode holder trace. A passing rerun is not a
diagnosis.
2. Rerun and land or close #280 (retire control and rooms TUI for
headless surfaces).
3. Record a two-machine north-star journey pass in
`docs/north-star-progress.md`.
Done means measured evidence: flake rate before and after over a stated
number of CI runs, #280 resolved, and a dated two-machine journey pass
recorded. Every later PR must state which metric it moved, with
before/after numbers. Private-relay and other test hardening, marketing,
and site work are paused; Dependabot and security fixes may continue.
<!-- CURSOR_AGENT_PR_BODY_END -->
<div><a
href="https://cursor.com/agents/bc-5203ab5a-0750-5412-a67a-37f2139b2317?cursor_ref=pr_footer&cursor_cta=open_in_web"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://cursor.com/assets/images/open-in-web-dark.png"><source
media="(prefers-color-scheme: light)"
srcset="https://cursor.com/assets/images/open-in-web-light.png"><img
alt="Open in Web" width="114" height="28"
src="https://cursor.com/assets/images/open-in-web-dark.png"></picture></a> <a
href="https://cursor.com/background-agent?bcId=bc-5203ab5a-0750-5412-a67a-37f2139b2317&cursor_ref=pr_footer&cursor_cta=open_in_cursor"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://cursor.com/assets/images/open-in-cursor-dark.png"><source
media="(prefers-color-scheme: light)"
srcset="https://cursor.com/assets/images/open-in-cursor-light.png"><img
alt="Open in Cursor" width="131" height="28"
src="https://cursor.com/assets/images/open-in-cursor-dark.png"></picture></a> </div>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: hraness <0thernet@users.noreply.github.com>
![https://cursor.com/assets/images/open-in-web-dark.png"><source](https://cursor.com/assets/images/open-in-web-dark.png"><source){kind=link}
![https://cursor.com/assets/images/open-in-web-light.png"><img](https://cursor.com/assets/images/open-in-web-light.png"><img){kind=link}
![https://cursor.com/assets/images/open-in-web-dark.png"></picture></a> <a](https://cursor.com/assets/images/open-in-web-dark.png"></picture></a> <a){kind=link}
![https://cursor.com/assets/images/open-in-cursor-dark.png"><source](https://cursor.com/assets/images/open-in-cursor-dark.png"><source){kind=link}
![https://cursor.com/assets/images/open-in-cursor-light.png"><img](https://cursor.com/assets/images/open-in-cursor-light.png"><img){kind=link}
![https://cursor.com/assets/images/open-in-cursor-dark.png"></picture></a> </div](https://cursor.com/assets/images/open-in-cursor-dark.png"></picture></a> </div){kind=link}
Scope and hold
Test/evidence-only PR; keep draft without auto-merge. Current head
90e361fe5ebc7e9846f723654e13c89372cbcc47changes onlycrates/vhalla-rooms-node/src/unix/{tests.rs,rejoin_trace.rs}. Bounded test observers record decoded sync requests/replies and verified validator peer keys, whitelist the restarted h3 WAL startup branch, and append only redacted numeric facts to the CI job summary after all recovery assertions pass. No runtime, WAL, journal, dependency, lockfile, or integration-owned plan change.Evidence and limits
Local Rust 1.98.1 locked/offline trace tests 6/6 and rejoin-plan test 1/1, strict all-target/all-feature rooms-node Clippy, workspace fmt, and focused diff check passed. Local socket bind returned EPERM; no local socket verdict. Exact-head Rust run 38632796561 passed
security,Requiredand rooms-node; the Linux job log reports crash/rejoin PASS (95.947 s) at90e361fe. Exact-head CodeQL run 38632784978 passed. Passing-test stdout was captured by nextest and does not appear in the job log; the bounded job-summary trace has not been independently read. Thus neither the observed WAL branch, proof-verified peer lifecycle nor the h3–h4 page exchange is established by these receipts. Do not promote on a green run alone. Earlier run 38628896097 failed a test-only WAL tracer that assumed a nonexistent node span; its failure log showed n3 receiving a request from n1 for h3–h4 and n1 receiving decoded [3,4]. That was not WAL or peer-proof qualification.Remaining acceptance
Obtain and independently inspect a bounded redacted socket-capable Linux trace (job summary or equivalent) and verify the observed WAL path (valid empty, replayed nonempty, or delayed replay superseded by a committed sync certificate), proof-verified source/victim peers, and source request / victim decoded-page height correspondence; verify stopped same-home n1 had no preloaded h3–h4 bodies and recovered h4 with preserved journal prefix, peer state and planned effects. Source/victim correspondence is by peer and height; libp2p inbound/outbound request IDs are local, not a shared wire ID. The original post-main restarted-h4 timeout remains unexplained even if this new run passes. Diagnose a demonstrated fault before any runtime change; rerun exact-tree gates and review before promotion. Never pad timeouts or rewrite/delete WAL or journal.