Three defects that all showed up as "the dashboard's numbers are wrong", plus
the notification and brand work asked for alongside them.
Tokens: Claude Code writes ONE JSONL line per content block of an assistant
message, and every one of those lines repeats the same message-level `usage`.
Both parsers summed them, so a tool-heavy session's tokens — and its cost — ran
1.5-2x high (one 4h session: $45.82 stored against $27.72 real). `lastUsageMsgId`
counts a message once, carried across the transcript cache's incremental chunk
boundaries. `replaceTokenUsage`'s high-water baselines mean the fixed parser
cannot correct what is already stored, so `setTokenUsage` (exact, baselines
cleared) is added for the deliberate re-parse path and `reconcileTokens` now
covers every session with a transcript on disk rather than imported ones only.
Turns: every `TurnDuration` row was stamped NOW while its dedup guard matched on
the transcript's timestamp, so each hook re-inserted the session's whole turn
history — 587k rows, 75% of a real events table, one session holding 68,747
copies of the same turns. The row now carries the transcript timestamp, which
fixes the guard and the feed's ordering together. Same class of bug in the
session metadata: `thinking_blocks` / `turn_count` / `total_turn_duration_ms`
are assigned from the full parse instead of `+=`, since `extract()` returns
whole-transcript totals on every call.
Events: leftover checkout hook entries in `~/.claude/settings.json` double every
event, every subagent row and every token figure. `stripLegacyHooks` has always
removed them — but only inside `bootstrap()`, which `fastPathOk` skips on
exactly the healthy installs where nothing else forces a full pass, so they
survived indefinitely. `hasLegacyHooks` (pure fs) now blocks the fast path.
A one-time repair at the first start after upgrading (`user_version` 3) deletes
the duplicated `TurnDuration` rows and re-parses every transcript with exact
writes. Marker 3 and not 2: an install that ran the earlier pass had its
corrected rows re-inflated by whatever old build was still ingesting hooks.
Router: switching a lane's AI Router while a pane is live only half-applies —
Claude Code re-applies `env` to a running session but keeps the model it
resolved at startup, so the old router's model name reaches the new endpoint and
the next prompt fails with "model not found". The apply route reports
`restart_required` / `run_id`, and the picker offers a Restart that respawns the
pane from the lane id alone (replaying the old `--model` would re-create the
mismatch).
Notifications (#29): the two moments Claude is BLOCKED on the user now have
their own switches and default to on — `onAwaitingInput` (a permission prompt or
a question) and `onTurnComplete` (end of turn). Before, the first was unswitchable
and the second hung off a preference that defaults to off, so the one thing a
user walks away from went unannounced. Prefs shape/defaults/storage move to
`client/src/lib/notifPrefs.ts`; Settings and the hook had private copies.
Brand: the sidebar shows the SmartGift wordmark with the AgentHub caption under
it, the collapsed rail shows the mark, and the favicon is vendored to match.
Closes#24.
**The review node ran nothing.** Both templates declared
`/superpowers:requesting-code-review` — the skill about how to HAND OVER a
review, not one that reviews anything — so `agenthub lanes node review`
performed no review at all. They now declare the bare `/code-review`, Claude
Code's own. It carries no `plugin:` prefix, so `requiredSkillPlugins` skips it
and the doctor's required-plugin row is unchanged (verified against the live
function: still exactly the five superpowers skills the templates name).
**The rest is the brainstorm → implement → review span, made cheaper to run and
harder to lose track of.** `ship-feature` gains four sections it never had:
- **Right-size the run.** The pipeline is not free — every dispatching phase
costs a model call and minutes, and the fix-loop pays it again each lap. A
typo does not enter the pipeline at all; a small single-layer change skips the
plan-debate, independent-test and gate agents and lets the user's acceptance
be the gate; medium/large runs everything. Sizing is a claim made out loud,
and a task that grows re-sizes upward.
- **Dispatching a judgement.** The fresh-context/FOREGROUND/never-`fork`/
hand-only-verifiable-artifacts/one-parseable-line contract, written once
instead of re-argued in three phases — plus the rule that was missing
entirely: never dispatch what a command answers. An agent is for judgement,
not for typing.
- **Resuming after a compaction.** Long runs outlive the context window; the
lane skill had this and this one did not. Re-derive position from the declared
stage and the diff, and treat a phase whose evidence you cannot find as one
that did not happen.
- **A `Leave when:` line on every phase.** Objective exit criteria are what stop
both the silent skip and the endless polish: intake leaves when the answers
are in hand, review leaves when the latest pass over the CURRENT diff is
clean, the gate leaves when verification really ran AND the user said yes.
And, narrower: intake asks only what the repo cannot answer (a question the code
could have answered spends the user's round and earns a worse answer);
implement works the approved plan task by task, re-checks each acceptance
criterion, heartbeats on long work, and treats a plan that turns out wrong as a
return to `plan` rather than a quiet substitution; tests runs the suite the diff
needs rather than both by reflex.
The `default` 8-node pipeline walked past its own intake: the node declared
`agenthub stage intake`, a stage report with no work attached, so "intake" was
a label the run passed through without ever asking a question. It now declares
`/superpowers:brainstorming` (and detects it — that rule moves off `plan`),
and both driving skills make the brainstorm mandatory, asked with
AskUserQuestion and waited on rather than answered on the user's behalf.
Three stops replace three silent slides:
- After the plan is written and adversarially challenged, it goes to the USER
for approval. Implementation does not start without it, and work that turns
out to need a different approach comes back for a fresh one.
- Review picks its reviewers from `git diff --name-only`: Claude Code's own
`code-review` always, a domain reviewer only for a layer the diff actually
touched, and — new — only once the change is big enough to have seams. Three
agents re-reading twenty lines buys one opinion at three times the tokens, on
every fix-loop lap. Review is also a loop now: it ends on a clean pass over
the current diff, not on one pass having happened.
- The gate asks the user to accept before anything is claimed done (and in the
lane pipeline, before the PR is published — publishing is outward-facing).
A NO-GO re-enters by SIZE: a defect goes back to implement, a wrong approach
back to plan. Nothing carries a NO-GO forward with a note about it.
Also here, from a real stall seen on a lane: `ship-feature` had no waiting rule
at all, and a session at the gate invented one with `ScheduleWakeup` — a
/loop-only primitive that paces nothing here and is refused outright when
`auto` mode cannot reach its safety classifier. The rule the lane skill already
had now exists in both.
Cost, measured in what each agent is actually for: the suite run leaves its
agent (nothing is judged, so nothing is delegated — an agent there buys a
second model's tokens to read output the session can read for free), and
`backend-reviewer` drops from opus to sonnet like its frontend twin; both read
a scoped diff against a fixed checklist, and the senior gate above them is
still opus.
Skill bugs found in the same sweep: doc-map told agents not to recreate the
removed translations and then, three lines later, to mirror every README change
into them; the docs mapping still carried `+VN+CN` and the punctuation left by
their removal; Procedure jumped 2 → 6; Verify required `scripts/doc-coverage.sh`,
which does not exist in this repo.
A lane's `remove` tore down what the lane pointed at: `git worktree remove
--force`, a branch delete, and a drop of the lane's own database. One
mis-click on a dashboard row cost a whole branch's unpushed code — and the
row is the only thing AgentHub actually owns. `remove` now stops the run and
the stack, clears AgentHub's own runtime bookkeeping under
`LANES_ROOT/.state/lane<slot>/`, and deletes the row. The working tree, its
branch, git's worktree records and the database stay exactly where they are,
for the user to delete by hand when they mean to.
That also fixes#28. Removal used to route through `assertDestroyable`, whose
third check demands the cwd appear in `git worktree list`; a managed lane
whose directory git no longer listed answered `ENOTWORKTREE` ("lane is not
listed as a worktree in …") and could never be dropped from the dashboard at
all. Removal no longer asks git anything, so no preflight blocker stops it,
and the unpushed-commits force-gate now applies to `reset` alone — nothing
`remove` does can discard a commit.
The UI follows: one label for both lane kinds ("Delete lane record"), copy
that says what survives, and no Force checkbox on removal.
`removeWorktree`/`removeLaneData` remain in place as guarded primitives; no
route calls them any more.
Closes#26, #25.
`runs` is the denominator every pass rate on the Workflows page is read
against, and any backwards move opened a new one. Stage detection bounces by
design — one feature reads `implement → tests → implement` as the session
edits, runs the suite and edits again — so a single feature was recorded as
five or six runs, and a lane in constant use reported a handful of passes
spread across dozens of phantom runs. A backwards move now opens a run only
when it is DECLARED; a backwards detection continues the run in flight.
Reaching the final node still closes the run for either, since work after
`done` is new work however it was observed. Replayed over this repo's own lane
the rule takes 26 recorded runs down to the 5 features that actually shipped
through it. The panel itself is unchanged: passes are still declarations only.
`/reload-plugins` and `/reload-skills` were typed back to back. Claude Code
opens its slash-command menu the moment `/` is typed and the Enter that follows
lands on that menu, so the second command was appended to the line still in the
composer and the pane submitted `/reload-plugins/reload-skills` — which Claude
Code runs as `/reload-plugins` with an argument, never reloading skills, and
leaving a stray prompt turn in the scrollback. Reproduced and fixed against a
live pane: SETTLE_MS (1s) between the two sends, which makes
`reloadOpenSessions` async rather than block the event loop, and the same pause
in the detached update-all worker.
The lane strip's tiles are now drag-reorderable (Alt+←/→ is the keyboard
path), and a lane's name can be set in place from the detail header. The
"Lane 23 ·" prefix is gone from both — the bare `#23` stays on the tile,
because that number is how the CLI addresses a lane.
Card order is a per-viewer preference, not a lane fact: it lives in
localStorage (`agenthub.workspace.laneOrder`) beside the split-view layout
and writes nothing server-side, so `GET /api/lanes` and `agenthub lanes`
keep answering in id order. Renaming is the opposite — it is PATCH
/api/lanes/:id with {title}, the one lane column a user now types into
directly.
That made a patch value reachable by hand for the first time, so
`bindable` now rejects any non-scalar for every PATCHABLE column with
EBADFIELD, and the route answers 400 for that and for the
SQLITE_CONSTRAINT family, rather than handing better-sqlite3 a datatype
throw and reporting 500 on plain bad input.
The uncommitted-files tooltip became a real panel. `GET /api/lanes/:id/git`
now returns `files` as {path, status, added, deleted} records — deltas from
one `git diff --numstat HEAD`, null where git has none to report (untracked,
binary, renamed) — and the card renders them portalled to the body, so the
Workspace's own scroll box cannot clip the list.
GET /api/lanes/:id/git now returns `files`, the raw `git status --porcelain`
lines the dirty/untracked counts were already parsed from, capped at 20 so a
lane mid-`npm install` cannot ship thousands of paths through a polled
endpoint. The lane card hovers them as the tooltip on its
"N modified · N untracked" text, appending "… +N" when the cap bit.
No new subprocess and no new endpoint — statusCounts was counting those lines
and throwing them away.
Four more spots that spent height saying what was already visible or
saying too much of it.
**LaneCard** drops the progress bar and percent entirely — the pipeline
map right below the card already draws the same number as named,
coloured nodes, and a bar beside it was a worse second telling of it.
**SplitLayoutToggle** (Workspace) swaps "1 pane" / "2 pane" / "4 pane"
text buttons for the conventional square / two-column / 2x2-grid icons
(`Square`, `Columns2`, `Grid2x2`) most apps use for this. The words move
to `aria-label` and `title` rather than disappearing.
**RunSetup** merges the fresh/resume source picker onto the same row as
the Working directory / Model / Permission mode / Thinking effort fields
instead of a second row above them — settings left, New/Resume session
right. The session picker for an active resume still gets its own row
below, since it only appears in that one state.
**LaneConsolePane**'s "Run Claude" header goes from a two-row title +
wrapped blurb to one row (smaller icon and title), and the blurb itself
is trimmed to one short sentence, hidden below `sm` where it would wrap
again anyway.
`docs/LANES.md` updated to match (split-view description, lane-card
field list, and `agenthub stage`'s progress-bar wording).
Verification: `tsc --noEmit` clean, `npx vitest run` 508/508 (2 stale
progress-bar assertions in LaneCard.test.tsx replaced with one asserting
the bar is gone), `npm run build` clean, screens.snapshot diff reviewed
before regenerating (icon buttons + new blurb + reordered DOM only), file
headers audit clean.
The Workspace lane panel spent four rows saying what fits in two.
Strip tile: the stage badges move onto the id row and the progress bar
and percent go away — the detail panel's pipeline map right below shows
the same progress with the stage names attached.
Lane detail: the pipeline picker, stage badges, feature picker, router
select and the remove button merge into the single header row instead of
a second row under it, and the pipeline's name is no longer interpolated
into `laneHeader` — the picker beside it already says it. The git block
collapses from four lines to one and drops the cwd, which the header
above it already spells out.
The router select is relabelled Model, moves out of LaneCard into that
header row (`LaneRouterRow` is now exported), and reports its apply
outcome as a ✓/⚠ marker with the message in a tooltip rather than a
wrapped sentence that pushes every row below it down.
Drops the tracker / dev QC / CI wait badge row and everything feeding it
(`useLaneIntegrations`, `api.lanes.integration`, the `integrations.*`
strings): two of the three toggles have no consumer at all, and a badge
reporting a toggle nothing acts on costs a row of height to say nothing.
`agenthub lanes integration` and the API route are untouched.
Two gaps left by the AI Router feature: a profile could be saved without ever
being reachable, and a lane's router could only be chosen at creation.
**POST /api/ai-routers/test** answers the question saving cannot. It sends one
Anthropic-style message capped at a single output token to `<baseUrl>/v1/messages`
with a 15s timeout — the wire shape a claude lane's eight env keys point at, so
what is tested is what will actually run. A router that answers is a 200 here
whatever it replies: `{ ok, status, message }` carries the router's own words,
and `status` is null when nothing answered at all. Only bad input is a 4xx.
The stored key is only ever sent to the endpoint it was stored against. A blank
`apiKey` falls back to the key of the profile named by `name`, but only while
the submitted `baseUrl` still equals the saved one — so testing a retyped
endpoint needs a retyped key (`EBADKEY`). Without that rule this route would be
a way to make the server POST a saved API key to any URL a browser names.
Network failures report `error.cause`, not `error.message`: node's fetch reports
every transport failure as the same useless "fetch failed" and hides the part
the user can act on (ECONNREFUSED, ENOTFOUND, a TLS error) one level down.
**The lane card carries an AI Router select.** The Add-lane modal was the only
place a router was chosen, and a lane outlives its creation. Switching goes
through `POST /api/lanes/:id/router/apply` — never a lane PATCH — because the
DB column and the lane's own config on disk may only move together. A failed
apply puts the select back rather than leaving the card claiming a router the
lane did not get. Hidden when `native` is the only profile: a select with one
option is a control that cannot do anything.
Adds `Lane.router_profile` to the client type (the payload already carried it).
A new **AI Router** tab in Settings, next to Codex, stores named router
endpoints (9Router, OpenRouter, a self-hosted gateway) as profiles: a base
URL, an API key, a default model and an optional per-tier override. The tab
only stores them; a lane CHOOSES one in the Add-lane modal, via a Native /
AI Router radio, and that choice is written into the lane's own config.
The two CLIs are configured differently, and the difference is theirs:
- a claude lane gets eight `env` keys merged into its own
`.claude/settings.local.json` (`.local` because the value holds a key);
- a codex lane gets `$CODEX_HOME/<name>.config.toml` — the layer
`codex -p <name>` loads — because Codex has NO project-level config.
Its base `config.toml` is never touched.
Native is a built-in profile rather than a code branch, so "go back to
native" is the same write path. It blanks the two connection variables
instead of removing them (an absent key at project scope lets a USER-level
router apply) and removes the model names instead of writing them (writing
them would override the user's own global model choice).
Key material never leaves the server: profiles live one file each under
`<CLAUDE_HOME>/agenthub-routers/` (0700/0600), `GET /api/ai-routers`
reports only `hasKey`, and there is deliberately no route returning one
profile in full.
Three things this had to get right, each found by review and verified
against the real tools rather than assumed:
- `wire_api = "chat"` is rejected outright by codex-cli 0.149 ("no longer
supported"), so every codex lane on a router would have failed to start.
The layer writes "responses", and writes `env_key` only when the profile
actually has a key — naming an unset variable is itself a startup error.
- The key must never be TYPED into a pane. `tmux.sendCommand` used
`send-keys` with a `KEY='…' cmd` prefix, which the dashboard streams to
every attached browser and bash appends to history; it now takes over the
idle pane with `respawn-pane -k -e`, passing env through argv the way
`newSession` always has. Verified on tmux 3.4: multi-element argv arrives
intact and the pane renders no trace of the value.
- Attaching COPIES a profile, so editing or deleting one in Settings has to
reach the lanes already using it. An edit re-applies it (a rotated key
would otherwise never reach the lane still holding the revoked one); a
delete resets those lanes to native and removes the Codex layer file.
Both report per-lane failures, which the tab surfaces.
`.claude/settings.local.json` is added to the lane's git exclude, for both
native and router lanes: `agenthub lanes reset` runs `git clean -fd` without
`-x`, so an un-excluded settings file is deleted on reset — and untracked
meant the key was committable.
DB: `lanes.router_profile`, nullable, additive. Written only by
`setRouterProfile` beside the disk write, never through `PATCHABLE`.
An optional checkbox in the Add-lane modal starts the new lane's terminal
run right after creation with `/sg-all:sg-md-writing` as its first message,
so the lane opens already writing or refreshing its CLAUDE.md/AGENTS.md set.
Claude lanes only — codex has no slash commands.
Lane Pipeline Pass Rates read `lanes.stages`, which is a snapshot of the
pass in flight: re-entering a node overwrites its record, and moving
backwards deletes every record ahead of the new stage. The report treated
one lane row as one run, so shipping ten features through a lane still
counted as one, and no node could ever exceed 1/1. History survived only
via `lane_features`, which is written when a feature is activated — a
path most users never take. The panel was not wrong, it was structurally
incapable of counting.
Add `lane_stage_events`, an append-only log written by every declaration
and every accepted detection, and aggregate from that. A run is one pass
through the template: the log starts a new one when the lane moves
backwards or leaves the template's last node. Existing snapshots are
seeded as one run each on first boot, so upgrading does not read as
"never run". Templates a lane is configured with are still listed at
zero — a pipeline going unused has to be visible as such.
Count inferred reaches too, in their own column and their own colour.
Detection stays out of the pass rate ("inference is never evidence"), but
reporting a stage the hooks demonstrably saw as "never reached" is what
made the panel look broken to anyone who works without running the
pipeline commands.
Also reject a stage naming no node of the lane's pipeline (400 EBADSTAGE,
listing the ones that exist) instead of storing it with a warning nobody
reads. A real lane had accumulated five such keys, one of them `test` on
a ship-feature lane whose CI node is `gates` — which is why that row read
never-reached while the stage had in fact been declared. Aliases still
resolve, so shorthand is unaffected.
Two leftovers from the rebrand, done together because they touch the same
docs and manifests.
**Data directory.** `agent-dashboard` could not simply be renamed in the
source: it names a real directory holding the live SQLite database. Startup
now moves it once, via `migrateLegacyDataDir()` in `server/lib/claude-home.js`
— a same-filesystem rename, so it is instant even at multiple gigabytes and
cannot leave two divergent copies. `getDataDir()` stays PURE and resolves to
the legacy path until the move happens; the first version of this put the
rename inside the getter, which meant `npm run test:server` relocated the
developer's real database as a side effect. The call sits behind
`require.main === module` and ahead of the router requires, because those pull
in db.js, which opens SQLite at module load. A failed rename is silent and the
old directory keeps being used: an unfashionable directory name beats a blank
database beside an unreachable one. Same treatment for the port-discovery file
(`.agent-dashboard.json` → `.agenthub.json`, old one dropped on write).
Not renamed on purpose: the Sagi localStorage keys (renaming silently resets
everyone's prefs) and the PWA manifest `id` (changing it orphans an installed
app).
**MCP server.** 4 tool calls in 46 days across 577,132 events, against 4,556
lines of TypeScript, a 2 MB committed build artifact, and a pre-commit guard
that had already blocked commits twice. Deleted: `mcp/`, the maintainer skills
and `mcp-reviewer` agent, the build check, `plugins/agenthub/.mcp.json`,
`docs/MCP.md`, 12 npm scripts, 5 Makefile targets, and every doc section
describing it.
Kept, because they are unrelated things sharing the name: the Config
Explorer's MCP tab and its API, per-lane Playwright MCP, the agenthub-config
plugin's MCP audit, and the historical `mcp__plugin_agenthub_*` events that
analytics must keep attributing.
Verified: server 1432/1432, client 459/459, tsc clean, prettier clean. Both
directions of the migration proven — a sandbox-HOME test run leaves the legacy
directory untouched, and a real boot moves it with its contents intact.
Clean break, no compatibility alias. Renames the product name, the CLI
binary (`ccam` -> `agenthub`), the env var prefix (`CCAM_` ->
`AGENTHUB_`), the Prometheus metric namespace (`ccam_*` ->
`agenthub_*`), the tmux session prefix (`ccam-lane-<id>` ->
`agenthub-lane-<id>`), the per-repo runtime profile directory
(`.ccam/profile` -> `.agenthub/profile`), the lanes state root
(`~/.claude/ccam-lanes` -> `~/.claude/agenthub-lanes`), the bundled
plugins (`plugins/ccam*` -> `plugins/agenthub*`) and their slash
commands (`/ccam-doctor` -> `/agenthub-doctor`, `/ccam:*` ->
`/agenthub:*`).
Because there is no alias, an existing install has to be updated by
hand: rename its `.ccam/profile` directory and any `CCAM_*` variables,
and reinstall the plugin under its new name.
Two other name families are deliberately left alone here, since they are
not just strings: the `Claude-Code-Monitor` repository/marketplace name,
and the `agent-dashboard` npm package plus its `~/.claude/agent-dashboard`
runtime directory, which holds the live database and needs a migration
rather than a rename.
Verified: server 1406/1406, client 459/459, `grep -i ccam` clean across
tracked files, `openapi.yaml` regenerated (95 paths, unchanged count).
A pipeline was all-or-nothing: `ccam ship-full` / `ship-quick` hand the whole
thing to a skill, and a node's own `command` — added so the map would stop
hiding the driver behind its detect regexes — was display-only. Re-running just
the CI gate, or redoing the brainstorm, meant clicking copy on the node and
pasting the string into a session by hand.
`ccam lanes node [<id>] [extra…] [--print]` runs it, in the lane's own cwd. A
slash-command node spawns `claude` with it as the prompt (same shape as
`ccam ship-full`); anything else is spawned directly. No id lists every node
and its command, which is also what the two error paths print.
The node is matched by id OR alias, the same way `ccam stage` matches one:
`default` calls the step `implement` and `ship-feature` calls it `implementing`,
and one command name must not depend on which template the lane happens to be
on. That needs the template's `aliases`, which the lane payload's
`pipeline_nodes` does not carry — so the two are merged, template for the
definition, lane for the `current` marker.
Both shipped templates now fill EVERY node's `command`. The seven that were
empty are the nodes whose only work IS the declaration — a terminal `done`, a
gate another agent declares — so they get `ccam stage <id>`, which keeps all 16
(and all 8) runnable instead of leaving holes a caller has to know about.
Slash commands for the same thing, because a session should not shell out to a
nested `claude`: `/intake`, `/plan`, `/implement`, `/tests`, `/e2e`,
`/e2e-passed`, `/review`, `/qc-plan`, `/qc`, `/gate`, `/publishing`, `/pr-open`,
`/reported`, `/watching-pr`, `/merged`, `/done`, plus `/ship-full` and
`/ship-quick`. Each resolves its node through `ccam lanes node <id> --print` and
runs the result IN the current session — invoking the slash command itself when
that is what the node names, otherwise running the shell argv. Resolving at call
time rather than baking the string into the file is what stops the 18 command
files from drifting when a template changes.
Two limits are deliberate:
- The command is split on whitespace and spawned WITHOUT a shell. A template
is data — anyone can drop a file into DASHBOARD_PIPELINES_DIR — so its
`command` must never be interpreted as shell. A node needing a pipe names a
script that contains one.
- Running a node does not declare its stage. A command executing is not proof
the node passed, same boundary as the console never writing a lane's stage.
Verified: npm run test:server (1403, +3 covering alias matching, full command
coverage of both templates, and every command file being listed in the
manifest), npm run test:client (459), prettier and check-headers clean. Docs:
docs/LANES.md § Running ONE node, docs/PLUGINS.md commands table, README +
README.vi CLI blocks, CLAUDE.md marker.
A pipeline template named ONE command for the whole run (`skill`), so the
only way to know what actually moves a given node was to read that node's
`detect` regexes. Each node now carries its own `command`: a slash command
(`/superpowers:writing-plans`) or a shell one (`ccam lanes hook ci-gate`),
whichever its detect rules already watch for.
The values are derived from each node's own detect rules, not invented —
every `ccam …` string was checked against bin/ccam.js and every
`/superpowers:*` against the installed plugin. Nodes reaching which is a
declaration rather than an action (`done`, `merged`, `reported`,
`e2e-feature-passed`, `qc-plan`) are deliberately left empty.
`pipelines.js` normalizes a missing or non-string value to "", so a
custom template written before this field keeps loading rather than
rendering "undefined". `nodeStates()` and GET /api/lanes/pipeline-stats
pass it through. Never executed by the dashboard — display text, the same
contract `skill` already had.
Surfaced where each view has room for it: the Workflows pass-rate panel
shows it under the node label and copies it on click; PipelineMap folds
it into the node tooltip, because a 16-node pipeline has to fit across
one card and there is no width for a second line.
`command?: string` is optional in the client type only to spare four
unrelated test fixtures — the server always sends the key, and the UI has
to branch on the empty string regardless.
Verified: npm run test:server (1373, +3), npm run test:client (459, +2),
tsc --noEmit, prettier and check-headers clean. docs/LANES.md § Custom
template format documents the field. openapi.yaml needs no change: the
lane payload and pipeline-stats were never schematized there.
A codex session emits no hooks and rarely calls `ccam stage` on its own, so
its card sat at `idle` forever while the amber "detected" badge was the only
thing that ever moved. `codex-ingest.js` now advances the declared `stage`
from its own written detection — the one place in CCAM where an inference
moves the blue `current` ring, and only for `codex` lanes.
It stays honest: gated on `recordDetection` having written (so forward-only,
declared-wins and the detection TTL are inherited unchanged), passes NO
evidence (the node renders `passed-no-evidence` amber, never `done`), and no
template's last node carries detect rules, so a lane can never be
auto-declared `done`. `setStage` grows a `keepDetection` option — its sole
caller — so the detection columns survive the advance and the card keeps
showing which command moved it instead of looking like an agent's own claim.
Two levers then get the evidence auto-advance cannot supply:
- `codex-nudge.js` now nudges on a stage held 2+ min with no evidence,
instead of on a detection sitting ahead of the declared stage — after
auto-advance the old condition can never fire, since the two now match.
- `ccam lanes agents install` on a codex lane writes the marker-delimited
`<!-- ccam:stage:start -->` block into the lane's own AGENTS.md — the one
instruction channel Codex reads — instead of copying agent templates into
`.claude/agents/`, which it never loads. Appended, user prose preserved,
reinstall replaces exactly that block.
Verified with a real `codex exec` run: stage moved to `watching-pr` with
`evidence: null` and the detection signal intact. 1364 server tests pass.
Three defects let a lane's pipeline map claim progress that was not this
pass's work, all of them on the "going back" path a new request takes:
- `setStage` cleared the detection columns on a real transition but never
pruned the `stages` records AHEAD of the new stage. A lane that ran through
`done` and was then declared back at `plan` rendered `plan` as `current`
while `tests`/`review`/`done` still read `done`, carrying the previous
pass's evidence, and contradicting `progress` (computed from the declared
index alone). A backwards declaration now drops every record it can prove
is ahead; a key matching no node is unplaceable and left alone.
- `activateFeature` rewrote the live lane row for the incoming feature but
carried the outgoing one's `detected_*` across. A never-seen slug starts at
`stage='idle'`/`stages={}`, exactly the shape `inferredDisplayStage` renders
from detection, so the new feature opened painted with the old one's
progress — and forward-only rejected every real detection for it until the
carried-over one aged past DETECTION_TTL_MS. Now cleared, like every other
real transition; deliberately not archived, since detection is live-only.
- `codex-ingest` classified a rollout journal by raw `now - mtime`, which goes
negative for a future mtime (clock stepped back, NTP correction, a restore
that preserved timestamps) and read as "written milliseconds ago". A
finished thread's journal could replay as live work; a file claiming to be
from the future is now idle.
Also fixes `check-headers.sh`, which audited via `find` and so reported local
gitignored runtime state (a lane's own `.ccam/profile/hooks/*.sh`) that is not
this repo's source and can never be made compliant here. It now lists files
through git (tracked + untracked-but-not-ignored), so a file an agent just
wrote is still caught before it is staged while ignored state is not.
Tests: 1359 pass (4 new). Header audit exits 0.
A codex lane's declared stage has always depended entirely on the session
reading AGENTS.md and calling `ccam stage` on its own — no hook, no wrapper,
no enforcement. Two additions make that more reliable without touching the
declared-wins / inferred-never-done invariants:
- LaneConsolePane pre-fills a codex pane's opening prompt with a plain-text
stage-declaration checklist built from the lane's own pipeline_nodes —
the same up-front priming a Claude lane gets from its pre-filled
/ccam:ship-feature-lane skill command, just as prose since Codex has no
slash commands.
- server/lib/codex-nudge.js sweeps every 60s and types a one-line reminder
into a lane's live tmux pane (tmux.sendText, a new literal-text primitive
alongside sendCommand) when its fresh detected_stage has sat undeclared
for over two minutes. It never writes stage/detected_stage itself — any
resulting declaration is still the codex session's own real ccam stage
call, just server-prompted instead of relying on memory alone.
server 1354/1354, client 449/449, tsc clean.
Two defects that made the 16-node ship-feature pipeline look like stage
detection was broken:
- stage-detect matched each rule against a JOIN of the tool input's
identifying fields. An Agent event really carries description + prompt +
subagent_type, so `^Explore$` (intake) and `^Plan$` (plan) could never
fire and both nodes shipped inert — the map jumped straight to
`implementing`. Rules are now tested one field at a time, which also
stops a rule matching across the seam between two unrelated fields.
- `detectionLeadsDeclaration` (bin/ccam.js) and `detectionLeads`
(LaneStripCard.tsx) located the declaration with `indexOf(lane.stage)`.
Most of what the skill declares is an ALIAS (`tests`, `booting`, `ship`,
`gate-blocked`), which is not a node id, so indexOf answered -1 and every
detection — however far behind — was badged as leading. Both now use the
node the server already resolved it onto (`state === "current"`).
server 1346/1346, client 447/447.
A lane on the 16-node `ship-feature` template showed an untouched pipeline
map and never declared a stage. Four independent causes, fixed at their
shared choke points:
1. Codex emits no hooks, so `touchLaneFromHook` — the only caller of
`stage-detect.detect()`, and correctly pinned to `claude` — could never
see a `codex` lane. New `server/lib/codex-ingest.js` tails Codex's own
rollout journals (`$CODEX_HOME/sessions/**/rollout-*.jsonl`), maps its
tool calls onto the hook-shaped `{tool_name, tool_input}` the matcher
already understands, and feeds the existing detect path plus an
idle/running status mirror under the same guard hooks use. Polled every
`DASHBOARD_CODEX_POLL_MS` ms (default 15000, `0` disables).
2. `resolveLaneArg` matched on cwd alone, so with a claude and a codex lane
at one directory (the composite `idx_lanes_cwd_tool`) a `ccam stage`
call could drive the sibling lane's card. It now resolves by
`--lane` → `CCAM_LANE_ID` → `--tool`/harness marker → owning lane, and
errors naming both candidates rather than picking one. Every pane the
dashboard starts carries `CCAM_LANE_ID` (`tmux new-session -e`, or an
assignment prefix when typing into a live shell pane).
3. `recordDetection`'s declared-wins veto had no exit: a lane left at the
pipeline's last node rejected every detection for the rest of its life.
A declaration on the LAST node no longer vetoes — the pipeline is over
there, so a later detection is new work. Nothing else is relaxed, and
detection still never writes `stage` nor renders `done`.
4. The run console pre-filled the pipeline's Claude slash command into
codex panes, which Codex reads as prose — so the skill that would have
called `ccam stage` never ran. Claude lanes only now.
A codex lane's *declared* stage still depends on the session calling
`ccam stage`; `AGENTS.md` is the only lever CCAM has, so it now says so.
Verified: server 1345/1345, client 441/441, and end-to-end against a real
codex lane row (detected_stage/detected_signal set, declared stage
untouched, no node rendered done).
Repo/Adopt mode (POST /api/lanes/ensure) treated lanes.cwd as globally
unique, so adopting a path already driven by one CLI silently ignored a
different tool choice. Uniqueness moves to the composite (cwd, tool):
one directory can now back a claude lane and a codex lane side by side.
- server/db.js: drop the column-level UNIQUE on cwd, migrate existing
DBs via table rebuild, add idx_lanes_cwd_tool.
- resolveLaneByCwd gains an optional tool filter; hook ingestion always
pins to "claude" since Codex has no hook mechanism here, so a hook
event never touches a codex lane sharing the path.
- POST /ensure resolves the existing owner by (cwd, tool) instead of
cwd alone, so a different tool creates a sibling lane.
- LaneConsolePane's cwd-only lane lookups now also match tool, so a
shared cwd doesn't attach a run to the wrong sibling lane.
Worktree mode is unaffected (already allocates a unique directory per
lane). docs/LANES.md and CLAUDE.md updated to describe the new
composite uniqueness.
The panel aggregated every lane at once, which answers "does this stage get
proved across the project" but never "does THIS lane prove it". A lane picker
now narrows the same aggregation through ?lane=<id> on
GET /api/lanes/pipeline-stats, counting that lane plus its own archived
features — dropping the features would hide most of a lane's history the moment
it starts its next one.
The response carries a `lanes` array that the filter never narrows: it
populates the picker, so it has to keep offering every other lane even while
the counts show one. For the same reason the picker renders above the empty
branch rather than inside the populated one — narrowing to a lane that has
recorded nothing would otherwise be a one-way trip, with no control left to
select anything else. A lane filter that is not a positive integer is rejected
with 400 EBADLANE rather than silently aggregating everything.
Restores the five analytics sections that turned out to be worth keeping
(patterns, model delegation, error propagation, session complexity, compaction
impact) and reworks what the page opens on.
The lane panel no longer mirrors the Workspace's live lane strip. It now
answers the question the strip cannot: how often each pipeline node actually
passes. GET /api/lanes/pipeline-stats aggregates every stage record of every
lane AND every archived feature — a lane's own `stages` only holds its current
feature, so without the archive the history resets each time a lane starts new
work. Evidence-backed passes are counted apart from bare claims, the same line
the pipeline map draws between `done` and `passed-no-evidence`.
Beside it, GET /api/workflows/skill-usage reports which skills and plugins the
sessions actually reach for. Only PreToolUse rows are counted: every call also
emits a PostToolUse, and counting both would double every number. A plugin's
total includes calls to the MCP tools it ships, so one used purely through MCP
does not read as unused.
Still dropped from issue #6: the agent collaboration network and the subagent
effectiveness scorecards.
A pipeline template may now declare the slash command that drives it
(`skill`), so the choice of pipeline is the choice of workflow rather
than just a picture on the dashboard:
- `ccam ship-full "<req>"` runs the 16-node `ship-feature` template's
skill (`/ccam:ship-feature-lane`) — through gates, QC and the senior
gate to a published PR, watched until a human merges it.
- `ccam ship-quick "<req>"` runs the 8-node `default` template's skill
(`/ccam:ship-feature`) — in-repo feature work that stops at the gate.
Each command puts the lane on its own template before spawning, so the
dashboard renders the right node count from the first stage instead of
after the skill self-corrects. `--print` shows the command, switches
nothing and runs nothing. The prompt reaches `claude` as one argv
element, never through a shell.
The same field feeds the Workspace run console, which pre-fills its
prompt with the lane's driving skill and never overwrites a prompt the
user has already edited.
Route Stage 4 / review-phase reviews to ccam:backend-reviewer,
ccam:frontend-reviewer, ccam:mcp-reviewer (parallel to the generic Code
Reviewer) by which of server/, client/, mcp/ the diff touches. Give each
reviewer a parseable verdict line and project memory to match qc-local's
rigor, and update docs/LANES.md's per-stage agent description accordingly.
Move .claude/agents/ and server/data/agent-templates/ship-feature-lane/
into one agents/ at the repo root — which IS the plugin root, so all five
agents now ship with the ccam plugin as ccam:<name>, the same way skills
already do.
The two pipeline agents were byte-identical copies in both places; that
duplication is gone. `ccam lanes agents install` now reads them straight
from agents/, so the per-lane copy in <lane>/.claude/agents/ still gives
the bare `subagent_type: qc-local` the ship-feature-lane skill prefers.
Skill and docs record both names: bare inside a lane, ccam: as fallback.
The repo root already IS the `ccam` plugin root (`"source": "./"`), so the
conventional `skills/` dir at the root is the plugin's skills dir — while
`.claude/skills/` only loaded for a session whose cwd was inside this
checkout. That is useless for a lane working in another repo, which is why
`ccam skills install` existed at all.
Move all seven skills to `skills/`: they now reach every plugin user as
`/ccam:<name>` (bare `/<name>` as an alias). `ccam skills install` stays as
the fallback for a plain checkout without the plugin, sourcing from the new
path.
Agents deliberately do NOT move. A plugin agent is addressed as
`ccam:qc-local`, which would break the `subagent_type: qc-local` references
in ship-feature-lane; `ccam lanes agents install` keeps writing them into the
lane's own `.claude/agents/`, the highest-priority agent source. MCP was
already plugin-shipped via `plugins/ccam/.mcp.json`.
Mechanical fallout: ~115 source files carry a doc-comment pointer to
`.claude/skills/file-headers/`, repointed to `skills/file-headers/`. The
generic Claude Code config-surface examples in openapi/cc-config (`.claude/
skills/code-reviewer`) describe any user's layout and are left alone.
Also fixes the two file-headers scripts, whose `ROOT=` climbed four levels
from `.claude/skills/file-headers/scripts/` and would have escaped the repo
from the new depth — it scanned sibling projects until corrected to three.
npm run test:server (1160 pass), test:client (422 pass), mcp:typecheck,
mcp:build + mcp:check-build all green.
The session that writes the code is the worst judge of it: it believes the
code is correct and remembers the reasoning its tests are meant to challenge.
Split the pipelines along a judge/do boundary — the driving session implements,
runs the mechanical gates, publishes and applies fixes; every stage whose job
is to judge the work is dispatched to a subagent with an EMPTY context.
ship-feature-lane:
- New hard rule: plan debate (0b), unit tests (1b), e2e (3), code review (4),
browser QC (6) and the senior gate (7) run in fresh-context agents.
`subagent_type: fork` is forbidden for these — a fork inherits the
implementing session's context, which is the bias being escaped.
- New Stage 1b: an independent agent writes the unit tests the implementation
is missing (test files only, no weakening existing tests) and returns
UNIT-TESTS: PASS/FAIL. A FAIL is fixed in the source, never in the test.
- Stage 3 delegates the e2e run + triage (the poll/timeout discipline moves
into the agent's brief); Stage 4 delegates the code-review skill run.
- Stages 6/7 keep their agents but may no longer be primed with the driving
session's context, conclusions or known-issue list.
- Stage 10: CI still red on the same job after a fix goes to a fresh agent —
calling your own failure a flake twice pushes a real defect past the gate.
- Fix-loop: a re-entry touching non-test source re-runs Stage 1b first.
ship-feature: same boundary for the plan, tests, review and gate phases.
docs/LANES.md updated to match.
Add detect rules for plan/review/gate/tests/intake nodes (Agent subagent
matches, TaskCreate, broader test-runner regex) so more of the pipeline
lights up without an explicit ccam stage call, and fix a review/gate
detect overlap where senior-gate-reviewer also matched the generic
reviewer pattern. Also fix TerminalView's TextDecoder dropping the
`stream: true` flag, which corrupted multi-byte UTF-8 chars split across
WebSocket frames.
`claude` only offers bypass in its shift+tab permission cycle when started
with `--allow-dangerously-skip-permissions`, which the pane argv never passed.
A dashboard-started run was therefore stuck cycling plan/auto/manual/
accept-edits, while the same session attached from a real terminal could
reach bypass.
The flag makes bypass SELECTABLE, not enabled — `claude --help`: "Enable
bypassing all permission checks as an option, without it being enabled by
default." The starting mode is still whatever `--permission-mode` says
(`acceptEdits` by default) and `--dangerously-skip-permissions`, which would
actually turn it on, is still never passed. Entering bypass remains an
explicit human shift+tab in the pane, or an explicit
`permissionMode: "bypassPermissions"` on POST /api/run.
Picking a session in the setup form's resume picker only staged the
selection: the user still had to type a prompt and press Run before the
lane's tmux session was started with `--resume`. A resume carries its own
transcript, so there was nothing to type.
The picker now fires the start directly with the picked session (passed
explicitly, since the parent's state has not landed on that tick), sends the
session's own cwd — which is what the locked cwd field already displays —
and the Run button no longer requires a prompt while a resume is selected.
Fresh runs still require one.
The cwd was synced only when laneId changed, but a pane can render before
GET /api/lanes has answered — split view restores its pane lanes from
localStorage, and layout 1 paints before the list loads. With no lane
resolved the field fell back to the home suggestion, and since laneId never
changed afterwards it stayed there. RunSetup submits that string verbatim to
POST /api/lanes/:id/start, so the run was started in the wrong folder.
Track the cwd in its own effect keyed on the resolved lane path rather than
on laneId, so it re-syncs as soon as the lane is known. The home default now
applies only while no lane is selected at all.
LaneConsolePane kept handle/cwd/prompt/runHistory in local state that was
never reset when the laneId prop changed, so selecting another lane swapped
the header and detail panel while the terminal stayed attached to the
previous lane's tmux session.
Reset the pane on lane switch and re-attach immediately to the new lane's
live run from GET /api/run when it has one, falling back to that lane's
setup form when it does not. `lanes`/`activeRuns` are read through a ref so
the page's 5s poll cannot wipe a half-typed prompt.
spawnRun adopted any existing `ccam-lane-<id>` tmux session without looking
at it, so a Resume issued while the session sat at a bare shell prompt (left
by `ccam lanes shell`, or by a `claude` that had already exited) dropped the
whole argv: no `--resume` ran, no initial prompt was typed, and the API still
answered 200. The lane's DB `run_id` is not set in that case, so the ERUNLIVE
guard in the start route never saw it either.
Reuse the pane instead of erroring: when the session exists and
`#{pane_current_command}` is a shell, type the argv into that pane and record
the run. A pane running a program (a live `claude`, an editor, a build) is
still adopted untouched, so attaching shows what is running rather than typing
over it. `sendCommand` POSIX single-quotes every argument and uses
`send-keys -l`, the only place in tmux.js that composes a command line.
The Workspace active-runs list only knew about runs this dashboard spawned,
so two `claude` sessions started by hand in terminal tabs showed up nowhere —
the list read "no active runs" while two agents were working.
Poll GET /api/sessions?status=active alongside the run list and merge those
sessions in as live rows, deduped against dashboard runs by session_id and
filtered to local sources with a cwd (a remote-source or cwd-less session
cannot be resumed on this machine).
External rows get no Attach action: the dashboard owns no tmux session for
them, so there is no PTY to bridge. They offer Resume, which reuses the
existing ensure-lane + start-with-resumeSessionId path to spawn a new
tmux-backed `claude --resume` in that folder — a second process on the same
transcript, not a view of the original terminal.
- Fix POST /api/run param: change 'prompt' to 'initialPrompt' (Finding 1)
- Fix POST /api/run response shape: show actual fields from publicRun() (Finding 2)
- Delete stale 'Running and releasing lanes' section with old POST /api/lanes/:id/start (Finding 3)
- Delete stale POST /api/lanes/:id/message section — endpoint now returns 400 EUNSUPPORTED (Finding 4)
All changes verified against actual code:
- server/routes/run.js:285 uses body.initialPrompt
- server/lib/pty-run.js:144-156 publicRun() returns {id, laneId, status, cwd, model, ...}
- server/routes/lanes.js:1134 message action returns 400 error
Lets a lane's pipeline template be switched live from the dashboard
(the same PATCH /api/lanes/:id the CLI's `ccam lanes pipeline` uses),
so lanes created before the picker shipped don't need the terminal.
Both ship-feature skills now force their own template before their
first `ccam stage` call, so the human never has to pick correctly at
lane creation.
Creation is the only point the UI could ever set a lane's template, and it
never offered the choice — so every lane added from "+ Add lane" was born
on `default` and rendered an 8-node map for a 16-node workflow, with no
screen able to change it afterwards. That is the defect that made the
ship-feature template unreachable from the browser.
The modal now shows a *Pipeline template* select fed by
`GET /api/lanes/pipelines`, labelled with each template's node count so the
consequence of the choice is visible. A failed fetch degrades to a `default`
option rather than blocking lane creation.
`pipeline` was already accepted by `POST /api/lanes` but silently dropped by
`/ensure` and `/worktree`, which build their own createLane payloads; both
now pass it through, and both map `EBADPIPELINE` to 400 like `EBADCWD`.
A lane's pipeline map only ever moved when a skill remembered to call
`ccam stage`, and the ship-feature template shipped with no detection rules
at all — so a lane driven by Superpowers skills sat at whatever stage it
last declared, and the `gates` node was never declared by anything.
Detection (`detect` rules on each node) now covers the Superpowers skill
invocations and the `ccam`/`gh` commands the ship-feature-lane skill
actually runs. It stays a safety net, not the mechanism: forward-only,
never `done`, never overriding a declaration. Two rules were deliberately
left out — `git diff` on `review` (this repo's own tests record it pinning
a lane at `review` on a real session) and anything on `merged`/`done`.
Stage vocabulary grows to 50 names over the same 16 nodes, following
Shipyard's PHASES shape: sub-states like `migration-collision`,
`e2e-scoped` and `gate-blocked` say WHY a lane sits on a node without the
map growing a node per reason. Every alias has a source — the skill
declares it, `default.json` uses it, or Shipyard's PHASES lists it.
Two silent failures fixed along the way:
- `lane.stages` is keyed by the raw declared string, so a stage declared
under an alias lost its `--evidence` and rendered amber instead of
green. `stageRecords` resolves each key onto its node.
- `ccam stage <typo>` stored fine and then rendered nowhere. It now warns
on stderr while still exiting 0.
`ccam lanes pipeline` closes the gap that made all of this invisible: a
lane could only be assigned a template at creation, and no screen in the
web UI offers the choice, so every lane added from "+ Add lane" was stuck
on `default`'s 8 nodes. An unknown template id is now refused rather than
silently falling back to `default` on read.
Also merges the repo's own `ship-feature` skill into the Superpowers
workflow: it delegates planning/TDD/review/verification instead of
restating them, and declares a stage at each phase.
- Repo mode adopts a directory as-is via /lanes/ensure (no worktree, no
branch fields) - the right choice for a main repo you want stage
detection on. Worktree mode (default) keeps the existing provisioning
flow but now requires a manually-typed branch name instead of deriving
one from the title.
- POST /lanes/worktree accepts an optional `branch`, validated via
`git check-ref-format --branch`; omitting it preserves the CLI's
existing auto-derived-branch behavior.
- New GET /lanes/browse lists a directory's immediate subdirectories,
backing a small folder-browse modal on both path fields - browsers
cannot expose an absolute path from a native picker, so this is
server-backed instead, consistent with the tool's local-first model.
Adopting the main repo as its own lane now gets stage detection
(cwd matches, same as any other lane), and its card lists every
managed-worktree lane provisioned from it with a jump-to link.
Also add the two missing Skill-tool detect rules (implement, ship)
so detection covers all four Superpowers workflow phases, not just
plan/review.
nntrivi2001