Repository navigation
feat: dl --refresh-stale recreates the stale workspaces that are idle - #676
Merged
Merged
Conversation
flows::refresh_stale::plan reads, per stale workspace, its launch lock, one docker exec of its process table, and the Claude sessions herdr's panes hold, and says Refresh or Skip with the reason. A build process, a shell under a Claude process, a busy or unresumable agent, a Claude session no pane holds, and a herdr that cannot be seen are each a skip. StaleImage now carries the container it was judged by, which is the container the plan reads. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
LaunchVerb::RecreateUnattended is recreate's path, agent sessions included, ending at Launched::Ready instead of a session. It is what dl --refresh-stale runs for each workspace, with nobody to attach. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
… seen dl --refresh-stale recreates outside herdr only a container it read no Claude process in, so the warning that the recreate ends the agents it cannot see was false there. A recreate typed by hand still warns. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
It prints one line per stale workspace (refresh, or skip and why), then runs the unattended recreate for each it may refresh, says each launch's notices as they happen, and ends with a count. A recreate that fails is counted and the loop goes on; the exit status is 1 when any failed. Named in dl.bash's global options and in the help's examples. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
…review/refresh-stale # Conflicts: # CHANGELOG.md # rust/devlaunch-core/src/flows/launch.rs
…sed a Claude outside herdr beside a held codex The rule knew only Claude processes. A codex running a command, or any codex at all outside herdr, left no Claude process to find, so the workspace was refreshed and the codex ended. And the count compared Claude processes against herdr's panes of every agent kind, so one held codex covered a Claude started outside herdr, which then ended with no line to start it again. An agent is now any agent dl knows by name (herdr::agent_named), found the way Claude was: by comm, or argv[0] or argv[1] for node. The shell rule and the count apply to all of them, so both sides of the count are agents of any kind. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
…rdr gives Seven cases the plan had no test for: a listing that ends halfway through a process, a stat line that does not parse, a held agent herdr does not answer for, a saved line in a pane herdr says holds no agent, a launch lock that cannot be asked, a Claude started under a Claude, and three stale workspaces whose verdicts must not leak into each other. Each was checked to fail with its branch broken. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
dl --refresh-stale judged every stale workspace once, printed the plan, and then recreated them in turn. A recreate can take minutes, so a workspace that was idle at plan time could have a build or a working agent by the time its turn came, and the recreate ended it. Each workspace the plan would refresh is now judged again just before its recreate. A skip then wins: it is counted as skipped and printed as "Skipping <ws> after all, judged again before its recreate: <why>". The plan is still printed whole, first. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
render_refresh_stale had no test. Four now run the binary against the read-side world with a docker fake that answers the process read: - nothing stale: the "nothing" line, exit 0, no docker exec. - one quiet and one with cc1plus: the whole plan is on stdout before the first devpod up, one recreate, refreshed 1 skipped 1 failed 0. - the first recreate fails: the second still runs, 1 failed, exit 1. - a stopped container: a Skipping line, no up, exit 0. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
…review/refresh-stale
pixi was on the build list, so a pixi shell that stays open in a pane kept that workspace stale for as long as it ran. A build that pixi run starts runs cargo, a compiler or another program the list already names, so the list loses nothing by dropping pixi. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
…a docker fault A stopped stale workspace was skipped as "could not read the processes in its container (docker exec failed ...)". Skipping is right, because a recreate would also start it, but the line said docker was broken. docker exec into a stopped container says "container <id> is not running", so the plan now skips that as Skip::Stopped, and dl says the workspace is stopped and that `dl <ws> recreate` starts it on the new image. (The Skip::Stopped variant and its match arm went in with the previous commit, which touched the same file.) Also: herdr's `done` is an idle agent nobody has looked at, so with the part 2 fix a done agent is held and a refresh may go ahead. The test that pinned the old skip now expects a refresh, and docs/cli.md says why pixi is not on the build list. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
Reviewer's GuideIntroduces Sequence diagram for unattended stale workspace refreshsequenceDiagram
participant User
participant CLI as dl CLI
participant Stale as stale_images
participant Planner as refresh_stale::plan
participant Docker as Docker container
participant Herdr as herdr
participant Launch as Launch
User->>CLI: --refresh-stale
CLI->>Stale: stale_images(workspaces)
Stale-->>CLI: stale images and container ids
CLI->>Planner: plan(stale workspaces)
Planner->>Docker: exec_as_root(read process table)
Docker-->>Planner: processes or unreadable
Planner->>Herdr: collect and read agent sessions
Herdr-->>Planner: pane states and saved lines
Planner-->>CLI: Refresh or Skip verdicts
CLI->>User: Print complete plan
loop Each planned refresh
CLI->>Planner: plan(workspace) again
Planner-->>CLI: Refresh or Skip verdict
alt Still eligible
CLI->>Launch: render_launch(RecreateUnattended)
Launch->>Launch: Recreate container
Launch->>Herdr: Restart held agent panes
Launch-->>CLI: Success or failure
else No longer eligible
CLI->>User: Skipping after all
end
end
CLI->>User: Print tally and exit status
Flow diagram for stale workspace eligibility checksflowchart TD
A[Stale workspace] --> B{Launch lock free?}
B -- No --> S1[Skip: another dl is launching it]
B -- Yes --> C[Read processes with docker exec as root]
C --> D{Container stopped?}
D -- Yes --> S2[Skip: it is stopped]
D -- No --> E{Processes readable?}
E -- No --> S3[Skip: processes unreadable]
E -- Yes --> F{Build process present?}
F -- Yes --> S4[Skip: build is running]
F -- No --> G{Agent has shell child?}
G -- Yes --> S5[Skip: agent is running a command]
G -- No --> H{Agent sessions present?}
H -- No --> R[Refresh]
H -- Yes --> I{Sessions visible and recoverable?}
I -- No --> S6[Skip: sessions unavailable or unresumable]
I -- Yes --> J{Every agent idle or done?}
J -- No --> S7[Skip: agent is busy or unanswered]
J -- Yes --> K{Running agents <= herdr panes?}
K -- No --> S8[Skip: sessions outside panes]
K -- Yes --> R
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
Codecov Report❌ Patch coverage is Additional details and impacted files
Flags with carried forward coverage won't be shown. Click here to find out more.
🚀 New features to boost your workflow:
|
Copied from the public-api CI job's diff. public-api.api.txt gains LaunchVerb::RecreateUnattended, a change to the promised API: a caller that matches LaunchVerb exhaustively needs an arm for it. Claude-Session: https://claude.ai/code/session_01Ro7QyyAtHmrh9oxyyK6h9m
…shape Found on a real host (#677, findings 1 and 2). devpod writes ContainerDetails.ID, not "Id", so the stale check flagged nothing and --ls and --refresh-stale never acted. herdr pane ids end in Crockford base32 (w3:pAX is pane 349), not decimal, so saved_argv failed for every real pane and a recreate ended the agent. The test fakes used the same wrong shapes; they now use the real ones.
…imes as text Found on a real host (#677, findings 6 and 7). With docker's containerd image store, a local image whose only tag moved is gone from docker image inspect while a container still runs it. A container made from the reference itself is now stale when its image id differs from the reference's id. Created is parsed as a timestamp, so a local build's +01:00 and a pull's Z compare correctly.
Found on a real host (#677, findings 3 and 5). Finding 5: with "features" in devcontainer.json, devpod builds a <prefix>:devpod-<hash> image whose hash covers the config, not the base, so up --recreate reused it and the workspace stayed stale. Before up, dl now runs docker image rm --force on that tag, which only untags while the old container runs. If that fails, recreate goes ahead with a notice and --refresh-stale skips the workspace. Finding 3: a line typed back into a pane got no DEVLAUNCH_HERDR_RESUME, so herdr kept only {"cwd"} and the second recreate ended the agent. recreate now reports the line it typed to herdr, retries while herdr says resume_not_accepted, and warns when herdr did not keep it.
blooop
commented
Oct 15, 2026
blooop
left a comment
blooop
left a comment
Owner
Author
There was a problem hiding this comment.
This was generated by AI during review.
wf-review of #676 at c94b5e7e. Fixed point: merge-base with feat/recreate-keeps-agent, 6b158582. Range: git diff 6b158582...HEAD, which has 18 commits, 20 files, +2628/−37. Two separate subagents ran the two axes in fresh context. Neither saw the code being written. CI was green when the review ran, except for review and gate.
Standards
Sources: AGENTS.md (CLAUDE.md holds the same text) and docs/development.md, then the Fowler baseline. I found no fixed-sleep retry in library code: wait is injected, and the threads are std::thread::scoped and joined.
rust/dl/src/render.rs:3677-3692: misplaced doc comment.derived_image_keptsits between the doc block oflaunch_refusaland thefn launch_refusal. Rustdoc now joins that block to the new helper's lines, solaunch_refusalloses its doc. Move the helper above the comment. Non-blocking. Fix it before merge.devlaunch-core/src/flows/refresh_stale.rs:263-327againstagent_worktrees/agent_lock.rs:681,763-790: Duplicated Code. This is a second/proc/<pid>/statparser and a secondstruct Process. The two parsers differ: one splits after the last), the other on") ". The repo rule says one source, or a diff test. Non-blocking.refresh_stale.rs:182: the flows layer classifies docker stderr.why.contains("is not running")matches the text of a formattedString. A typed "not running" error fromclients::dockerkeeps a change to the message from turningStoppedintoProcessesUnread. Non-blocking.flows/launch.rs:5217-5234: Repeated Switches and bool flags.matches!(verb, Recreate | RecreateUnattended)appears twice, and two morematches!become the bare boolsunattendedandwarn_unseen. Addis_recreate()andis_unattended()toLaunchVerb, besiderebuild(). Non-blocking.launch.rs:2998: a panic is swallowed.report.join().ok().flatten()treats a panicked report thread as "line kept", so the user seesAgentSessionRestartedand notAgentSessionLineNotKept. MapErrtoSome(pane_id). Non-blocking.launch.rs:2975,:3022: Duplicated Code. Theseqandreadclosures are copied frombeside_resume_report. Non-blocking.dl/src/commands.rs:1286-1371: Divergent Change. The binary holds part of the refresh policy: it judges again before each recreate, lets a skip win, and counts the tallies. The core module says it "decides and does nothing". Onlyread_side.rsend-to-end tests cover that half. Non-blocking.- Minor: Data Clump and Primitive Obsession.
LaunchNotice::DerivedImageKeptandLaunchRefusal::DerivedImageKeptrepeat{workspace_id, image, why}.Planned.workspace_idis aString. Both follow the existingStaleImagesconvention. Non-blocking.
Axis verdict: Comment. Nothing is blocking.
Spec
Spec: #673, the body of #676, and the decision comments on #677 (6084475333, 6128046327).
The spec items below match the code:
- Stale and idle only. #673: "applies 2 to each stale workspace whose Claude sessions are all idle … and that runs no build; it skips busy ones and reports them".
planlooks only at stale workspaces (refresh_stale.rs:137). It checks builds (:185), an agent with a shell child (:199), idle or done state (:240), and agents that no herdr pane holds (:253). Each skip line gives its reason. The code judges each workspace again just before its recreate (commands.rs:1331). - Exit codes. A failed recreate counts as a failure, and a skip does not (
commands.rs:1366).Ending::Refusedgives exit 1. - Finding 3, option (a). "the recreate reports the same line to herdr … retries on
resume_not_accepted(1 s, up to 20 tries), one thread per pane". The code does this atlaunch.rs:2957,agent_sessions.rs:286, andsession_manager.rs:480-481. The noticeAgentSessionLineNotKeptis atlaunch.rs:5380. - Finding 5, options (a) and (b) as built. "runs
docker image rm --force <tag>… If that fails:recreategoes ahead with noticeDerivedImageKept;--refresh-stalerefuses that workspace". The code does this atdocker.rs:98-101andlaunch.rs:5270-5308, andstale_images.rs:201-203checks that the tag differs from the reference. - Finding 4 ("The typed line's argv[0] is
dl") is not fixed, and the PR says so. I do not count it against the PR.
Findings:
- "reads
session.jsonback. When herdr does not keep the line, dl prints a new noticeAgentSessionLineNotKept".report_againtreatsResumeReported::TakenasKept(agent_sessions.rs:306).report_resumealso returnsTakenwhen it cannot readsession.json. In that case dl never confirmed the line, the notice does not show, and the next recreate can end the agent. Non-blocking. - Option (b): "
--refresh-staleskips such a workspace and says that a recreate cannot move it". The plan still prints "Refreshing " for that workspace. The refusal comes later, when the untag fails (render.rs:3714). The result is correct, but the plan line for that workspace is wrong. Non-blocking. - #673 "Done when": "
dl --refresh-stalerefreshes idle stale workspaces and leaves busy ones alone". The PR body said "I did not run: A real refresh." Non-blocking. Note from the run owner: the host run on 2026-10-15 atc94b5e7e's tree did run a real refresh. Two refreshes and two recreates with an idle agent passed, and the run found no stale workspace left behind. The evidence is in the "Found on a real host" section of the PR body.
Axis verdict: Approve. Nothing is blocking.
Verdict
Comment. Neither axis has a blocking finding. Standards 1, the misplaced doc comment, should be fixed before merge. The other findings can be follow-ups.
Found in the wf-review of #676 (Standards 1). The helper sat between launch_refusal's doc block and the fn, so rustdoc joined the two and launch_refusal lost its doc.
Found in the wf-review of #675 (Standards 1). agent_sessions and docs/cli.md treat done as an idle agent nobody has looked at; the enum doc said the agent had ended, which invites dropping done panes.
The saved resume line herdr types back always began with a plain `dl`, so a pane started with the dev build came back under the released one. The line now names the dl this process was started as: argv[0]'s file name, with aid's name mapped to its dl, falling back to `dl`. Not current_exe(), which resolves pixi global's trampoline to a path that moves on upgrade. One function, herdr::restore_program, feeds both the reported line and the hook's prefix.
report_typed_lines_again joined each report thread with .ok().flatten(), so a thread that panicked counted as kept and the user saw no notice that the next recreate would end the agent. A panic confirms nothing, so the pane now counts as not kept. The per-pane threads move into lines_not_kept, which takes the report as a closure so a test can make it panic.
report_again mapped ResumeReported::Taken to LineKept::Kept, and report_resume returns Taken when it cannot read session.json back, so a line nobody confirmed went by as a restart that went well. It now maps to LineKept::Unchecked, and the launch says so with a new notice, AgentSessionLineUnchecked: herdr took the line and dl could not check that it kept it. That is not the same as herdr refusing it, so it does not reuse the not-kept wording.
…aunch beside_resume_report and report_typed_lines_again each built the same wall-clock seq closure and the same read closure, and the agent-session collect built the read a third time. They now share report_seq and read_session_file.
restart and the not-restarted notice each joined HeldSession::line into a shell line by hand. HeldSession::shell_line now does it once.
refresh_stale told a stopped container from a docker fault by searching the error words in the flows layer, far from the docker call that wrote them. exec_as_root now reads docker's stderr itself and returns ExecFailure::NotRunning, and refresh_stale matches on that variant.
agent_lock and refresh_stale each parsed the stat line their own way. The two agree on every real line today, but two copies are two chances to drift. domain::proc_stat now holds the one parser, which ends the command name at the last ')', and both flows read through it. Test fixtures in refresh_stale now print whole stat lines, as the kernel does.
…attended run_rebuild asked matches!(verb, Recreate | RecreateUnattended) twice and passed drop_derived_image and hold_agent_sessions a bare bool each, one named unattended and one warn_unseen that was its negation. The verb now answers is_recreate() and is_unattended() itself and the two helpers take the verb, so the call sites read as what they ask.
restart_agent_sessions took up: bool and did one of two unrelated things on it: start each agent in its pane, or hand each line to the user. The caller now picks by the up outcome, and each half has its own name.
…n claude --resume The notices for agents a recreate ends and cannot restart named 'claude --resume' as the way back, and render.rs picked claude's words with agent == "claude". A codex or gemini pane got either no resume or claude's. Core now keeps each known agent's resume words beside AGENT_NAMES (herdr::resume_line / resume_lines, exported for dl), a Python test diffs them against aid's AGENTS table, and the notices name the agent's own resume, or every known agent's when dl cannot see which one ran.
Only collect and refresh_stale call it, both inside devlaunch-core, so it leaves the public API. Drops its public-api.rest.txt row.
It folded AgentReading into an Option<Option<String>> and then matched that, so a reader had to decode what None and Some(None) meant. The match now names each reading, and every case is spelled out.
agent_sessions::Manager::resolve was a copy of ResumeReport::resolve: the provisioning gate, session.json beside the socket, and the PATH fallback for the binary. herdr::Locator now holds that rule once; the pane id is optional in it because the scan needs none and the report requires one.
Manager sat beside flows::session_manager and said nothing about what it holds: the herdr whose panes a recreate scans and types lines into. The public-api.rest.txt rows move with the name; their order is unchanged.
The binary judged each workspace again before its recreate, let a skip found then win, and counted the tally. That is refresh policy, so it now lives in refresh_stale::run, which takes the recreate through a Refresher and returns a Tally. Core tests cover a workspace that turns busy after the plan and the tally. The binary keeps the wiring and the lines.
LaunchNotice::DerivedImageKept and LaunchRefusal::DerivedImageKept both spelled out workspace_id, image and why. They now hold one struct, so the two cannot drift apart.
The rows for refresh_stale's Refresher trait were added by hand above the BUILD_PROGRAMS row, where cargo public-api renders them below it. CI's public-api job diffs the order too.
…feat/refresh-stale
main made dla a second name for aid (#679), so restore_program maps dla and dla-next the way it maps aid and aid-next.
This was referenced Oct 16, 2026
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Part of #673 — A workspace keeps its old image after a pull, and a recreate ends its agent. This PR is part 3 of 3:
dl --refresh-stalerecreates each stale workspace whose agents are idle. It can run after the daily pull.What this changes
dl --refresh-stale. It takes no workspace, no-yand no--force. It is in the README command table,dl.bashglobal_opts, the--helpexamples anddocs/cli.md.flows/refresh_stale.rs, makes the plan. It changes nothing. For each stale workspace from part 1, it gives "refresh", or "skip" with the first rule that fails.LaunchVerb::RecreateUnattended, runs the recreate path of part 2 with no attach. Part 2 then starts the agents again in their herdr panes.StaleImage::container()gives the container id that part 1 already reads.rust/dl/src/render.rs.Where this fits
"Agent" means each agent that
dlknows by name:claude,codexandgemini, run directly or bynode. herdr'sdonemeans "idle, and nobody has looked at it yet", soidleanddoneboth let a refresh go ahead.How I checked it
All tests run on
FakeRunner, or on the real binary with a fakedockeronPATH. This devcontainer has no herdr and no docker daemon.node, a process list that ends early, a stat line that does not parse, herdr that does not answer, and several workspaces with different verdicts. I broke three rules on purpose, and five tests failed.rust/dl/tests/read_side.rs): nothing stale, the plan printed before any recreate, exit 1 with the loop going on after a failed recreate, and a stopped workspace skipped as "it is stopped".cargo test --workspace, clippy with-D warnings,cargo fmt --check: clean.pytest test/against the debug build: 850 passed.prek run --from-ref origin/main: passed.I did not run:
A real refresh.Done on 2026-10-15 on a real host: see "Found on a real host" below.scripts/public-api-snapshots.sh. The tools are not in the container. I copied the diff from thepublic-apiCI job into the snapshot files.Merge danger
Door: two-way. The flag is new, and nothing else calls the new verb. The promised API (
public-api.api.txt) gainsLaunchVerb::RecreateUnattended, so a caller with an exhaustivematchonLaunchVerbneeds an arm for it.Blast radius: medium. A false "refresh" ends a working agent. Each rule therefore skips when it is in doubt. These limits remain:
dl <ws> recreate.docs/cli.mdsays this.cmdlinewords, to end in an agent name. A native install that starts a versioned binary by its own path is not seen as an agent. If it runs outside herdr, a refresh ends it.Found on a real host
A test on a real host (docker 29 with the containerd image store, devpod v0.26.3, herdr 0.9.5) found seven defects that the unit tests missed, because the fakes used the same wrong shapes. Details are in #677. Commits 34ec914, 3fa0ece and c94b5e7 fix six of them:
ContainerDetails.ID, not"Id". The stale check flagged nothing, so--lsand--refresh-stalenever acted. Fixed in 34ec914; the fakes now say"ID".w3:pAXis pane 349).saved_argvparsed them as decimal, so a recreate ended every agent. Fixed in 34ec914; the tests use base32 ids.resume_not_accepted. Fixed in c94b5e7.dl, so a resumed session runs whicheverdlis on PATH. Not fixed: it matters only when a dev build runs beside the released one. The host test put adlshim on PATH in the agent pane.<prefix>:devpod-<hash>image when devcontainer.json has"features", so the workspace stayed stale. dl now runsdocker image rm --force <tag>beforeup, which only untags. Fixed in c94b5e7.docker image inspect. Fixed in 3fa0ece: a container made from the reference is stale when its image id differs.Createdwas compared as text, so+01:00andZsorted wrong. Fixed in 3fa0ece; it is parsed as a timestamp now.Host check on the tested tree (all pass):
dl --lsprintedproj runs an older image than dl673-test:latest now names.dl proj recreateprintedRemoved the tag proj-736a5:devpod-1c27…3834 that devpod built it under, so the recreate builds the image again from the newer one.docker events show this order: untag, new image built, old container stopped, new container started. docker gave no error./stale-markeris in the new container, and--lsno longer says stale.docker image rm --force <tag>prints onlyUntagged:and exits 0. Both containers keep running.Started the agent in pane wG:p4P again.herdr's session.json kept--resume <id>both times. After the second recreate, the agent remembered the word from before the first.dl --refresh-stalerun twice, each time after a new stale build. Both runs printedRefreshing proj…andRefreshed 1 stale workspace, skipped 0, and 0 failed.and exited 0. With nothing stale, it printedNo workspace runs an older image than its reference names.and exited 0.--lsstill named the workspace stale after the old base left the store.Unit tests on that host: 2232 pass. Four "pushed by URL"
flows::lifecycletests fail, and they fail the same way on f5551ce without these commits. CI'srustjob passes them.Stack
🤖 Generated with Claude Code
Summary by Sourcery
Add an unattended stale-workspace refresh workflow that safely recreates eligible workspaces, restores their agents, and reports skipped or failed workspaces.
New Features:
dl --refresh-stalecommand to recreate stale workspaces whose agents are idle and whose containers have no active work.Bug Fixes:
Enhancements:
Documentation:
Tests:
Follow-up fixes
From the review in #677. One commit per finding.
aid-nextmaps todl-next, falls back todl), so adl-nextpane resumes underdl-next. Notcurrent_exe(), because of pixi global's trampoline. A path is cut to its file name because herdr refuses a path.AgentSessionLineUnchecked, not silence.clients::dockerreturnsExecFailure::NotRunning;refresh_stalematches on it, not on stderr text./proc/<pid>/statparser (domain::proc_stat, last-)rule), with a test for a comm that holds). The twoProcessrecords stay apart because each holds a field the other flow never reads.LaunchVerb::is_recreate()/is_unattended()replace thematches!checks and theunattended/warn_unseenbools.report_seq()andread_session_file()replace the copied closures.herdr::Locatorfor the resume report and the pane scan.collectmatches onAgentReadingdirectly.AgentStatus/PaneAgent::status(refresh_stale::sessions_skipgivesSkip::AgentBusyfrom it).HeldSession::shell_line().restart_agent_sessionsandhand_back_agent_sessionsreplace theup: bool.herdr::AGENT_RESUMES), notclaude --resumefor every agent. A Python test diffs that table against aid's.agent_readingispub(crate).agent_sessions::Manageris nowHerdrPanes.refresh_stale::run, with unit tests there.DerivedImageKeptstruct for the notice and the refusal.Public API. The promised API (
public-api.api.txt) changes:LaunchNoticegainsAgentSessionLineUnchecked { pane_id }, andLaunchNotice::DerivedImageKeptnow holds the structflows::launch::DerivedImageKeptin place of three fields. A caller that builds or destructures that variant needs a change.DerivedImageKeptis not re-exported fromapi, so the residual count goes from 40 to 41 types. Inpublic-api.rest.txt:HeldSession::shell_line,herdr_resume_line(s),refresh_stale::{run, Refresher, Tally}and theDerivedImageKeptstruct are new;LaunchRefusal::DerivedImageKeptholds the struct;agent_readingis gone;Managerrows are renamed toHerdrPanes.