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.
withDetected() rendered lane.detected_stage forever once written — the TTL in
recordDetection only stops future writes from advancing past a stale floor, it
never clears the row, so a lane gone quiet kept painting amber nodes from
activity that was no longer current. Also add the shared test-runner regex to
the ship-feature pipeline's `gates` node: Stage 1b's independent-test subagent
runs the repo's test command directly, not through `ccam`, and had no
detection fallback if `ccam stage tests` was skipped.
lane-agents.test.js now drives installAgents through real lane records
(lanesLib.createLane) instead of ad-hoc git fixtures, and asserts the
agent-template source moved to agents/ (not the retired
data/agent-templates/ path). New skills-install.test.js covers the
ship-feature-lane skill copy resolving from skills/, not .claude/skills/.
Unmatched Ask questions now adopt a dedicated "Sagi" lane (POST
/api/lanes/ensure, cwd defaults to $HOME) and spawn a tmux-backed claude
run there instead of just prefilling the Run page composer.
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.
Move the rare destructive lane actions behind a LaneToolsMenu next to
the lane header, and show a liveness dot + status label + stage age
inline in the header instead of only in the card grid.
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.
SAGI_SIZE (the draggable button's real footprint) and SagiAvatar's default
`size` both go 60 -> 90, and the viewBox drops the exporter's padded
`0 0 1280 1280` canvas for a tight window on the art: the measured bbox over
every path is x 331..1114 / y 173..1044, padded out to a square and raised to
y 106 so the zzz / bang / sparkle mood overlays stay inside and the box keeps
the 1:1 aspect of the square width/height.
Path data is not rescaled, so MAX_PUPIL_SHIFT and the eye/mood
transform-origins in sagi.css stay valid in the original 1280-unit space.
`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.
The previous `attachCustomWheelEventHandler(() => false)` killed scrolling
outright. xterm's `bindMouse` consults that handler in BOTH wheel paths — the
one that emits a real SGR mouse report and the fallback that turns a notch on
an alt-screen buffer into ESC[A/ESC[B — so a blanket false also blocked tmux's
own wheel scrolling (tmux runs with `mouse on`, i.e. tracking is active).
Now the handler returns false only for the exact bad case: no wheel tracking
AND an alternate buffer. `x10` counts as untracked because that protocol
reports button presses only, never the wheel, so xterm takes the emulation
path for it too.
closeWebSocket() nulls both WebSocketServer references but cannot unregister
the `server.on("upgrade")` listeners that initWebSocket/initPtyWebSocket
installed on the shared http.Server. An upgrade landing in that window threw
TypeError: Cannot read properties of null (reading 'handleUpgrade')
from the listener — unhandled, so it killed the process mid-SIGTERM instead of
letting it exit gracefully, and no server came back up. Observed in
runtime/server.log right after a restart, with the dashboard then showing
"Mất kết nối" and nothing else.
The previous commit's TerminalView reconnect makes this near-certain rather
than rare: every open run console re-attaches to /ws-pty every 1.5s, so a
shutdown almost always has an upgrade in flight. Both listeners now destroy
the socket when their server is gone; clients retry, which is the correct
answer during a shutdown.
Regression test uses a real http.Server and raw upgrade requests — without the
guard the process dies and the test run hangs rather than reporting a failure.
xterm emits ESC[Z for Shift+Tab but, unlike plain Tab, never marks the
keyboard event cancelled, so the browser ran its default action and moved
focus out of the terminal. The pane saw only the first press, which made
`shift+tab to cycle` (claude's permission mode) look dead on /run while
working fine in a real terminal.
A custom key handler now calls preventDefault() on that one combo and still
returns true, so xterm's own handling — including the ESC[Z it sends — is
untouched; only the focus-moving default is suppressed. Verified against a
scratch tmux session: writing ESC[Z into the attach pty cycles claude's
status line from `bypass permissions on` to `auto mode on`.
Trade-off: Shift+Tab no longer tab-reverses out of the terminal; click
elsewhere to move focus.
Three separate reasons the Workspace terminal looked unscrollable:
- xterm falls back, on an alt-screen buffer whose program has not enabled
mouse tracking, to translating each wheel notch into a cursor-key press.
Inside a `claude` pane that reads as arrow up/down, so scrolling walked the
prompt history. A custom wheel handler returning false suppresses only that
emulation branch; real mouse reports come from a separate listener xterm
registers when the program does request wheel events.
- TerminalView never reconnected, so a server restart left the pane frozen on
its last painted frame — still live-looking, but swallowing every keystroke
and mouse report. It now re-attaches 1.5s after an unexpected close and
stops only on the server's `exit` frame; tmux keeps the session, so the
re-attach repaints in full.
- tmux repaints the whole pane, so xterm's own scrollback is always empty and
its viewport scrollbar renders as a groove with no thumb that cannot move.
Hidden, since scrolling there goes through mouse reports, not the DOM.
Sessions created before the mouse-on fix landed (or before the server
restarted to pick it up) never got the set-option call, so their pane
still had no drag/wheel scroll. Moving it into pty-attach's attach()
makes it idempotent per-connection instead of once-at-birth.
Detail toggle was hidden and info forced-visible whenever layout was
single-pane; user needed the ability to hide it too. Scrollbar track
color matched the page background (surface-0), making it invisible —
moved it to surface-2 so the track reads against any surface.
Full rename across files, identifiers, CSS classes, and user-facing strings
(en/vi locales, ARCHITECTURE.md). Replaces the hand-drawn animated cat
avatar with the pig-superhero mascot artwork, split into parts to keep
per-eye cursor tracking and blink animation working; whole-mascot mood
transforms (breathe/bob/shake/tilt) and new zzz/bang/sparkle overlays
carry the rest of the mood expression since the traced art has no shared
palette to key off of.
Claude Code spawns plugin MCP servers straight from the read-only
git-content plugin cache, which never gets an npm install for mcp/. The
old tsc-multi-file build imported bare specifiers that only resolved
via mcp/node_modules, so every fresh plugin install/update produced a
-32000 reconnect failure. Bundle mcp/src into a single dependency-free
build/index.js instead; the one CJS-heavy import (express, only used
by --transport=http) is now dynamic so the stdio path never evaluates it.
The lane strip was a fixed-width vertical column beside the detail panel.
Flipped it to a horizontally scrolling row on top so the detail/console
area gets the full page width below it.
Session panes had a scrollbar track but no way to drag it — wheel/drag
events passed straight to the running program instead of entering tmux's
copy-mode scrollback. xterm.js already forwards mouse-tracking escapes
once tmux advertises them, so turning mouse mode on per-session is enough.
Docs and the command's own instructions previously implied /ccam-update
fetches new code; it only reinstalled deps and restarted the server
against whatever version was already resolved. It now also runs
`claude plugin marketplace update` + `claude plugin update ccam@<mp> -y`
first, and says plainly that /reload-plugins is still a manual step
(no CLI/script equivalent exists to apply it automatically).
The lane-switch effect only synced `handle` from whatever activeRuns
snapshot was already loaded, so navigating away and back to /run (a
fresh mount with activeRuns still null) left a running lane's console
stuck on the setup form until the user switched lanes and back.
Adds POST /api/updates/apply (pull, rebuild, restart) plus an "Update now"
button and hourly auto-check in the UI, for checkouts that are safely
fast-forwardable. Reintroduces self-restart (previously removed in edc25ca
for cross-environment reliability concerns) per explicit user request.
Without a lane, the picker had no cwd filter and listed every session
across every repo. Fall back to the free-typed cwd field so resume
suggestions stay scoped to the working directory in view.
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.
Four cases in lane-lifecycle.test.js call /start without stubbing PATH,
so they spawn the real system `claude` binary in a real tmux session
to simulate a stuck/live run. Each then mocks tmux's own exec calls to
fake has-session/kill-session for the app's checks, but never touches
the real spawned process — the mock only fools the app, not the OS.
Two of these leaked past every prior test run undetected (ccam-lane-22,
ccam-lane-24), surfacing in the dashboard's live "Dashboard runs" list
with no DB record and a garbage started_at, and reappearing in a
Workspace split pane pointed at a deleted temp directory.
Stub a lightweight fake `claude` on PATH (same pattern already used
correctly elsewhere in this file) instead of spawning the real CLI, and
explicitly kill the real tmux session in each test's teardown since the
app-level mock never reaches the OS process.
The lane-detail section is a flex-1 row item next to the lane list, but
lacked min-w-0. A flex item's intrinsic min-width defaults to its
content size, so PipelineMap's 16-node row (which relies on flex-1
min-w-0 truncate per node to shrink) pushed the whole panel wider than
its allotted space instead of compressing, spilling nodes off-screen.
Undeclared lanes default stage to the DB sentinel "idle", which never
matches a pipeline node — skip the mismatch warning in that case
instead of showing a false lane-action-failed banner. Also cap the
expandable lane-info block so it can't squeeze the console/split-view
out of the fixed-height detail panel.
The detail panel (LaneCard/PipelineMap/proof gallery/console) could
grow tall enough to visually crowd out the lane-strip carousel above
it once a run was attached in split view. Move the lane list into its
own scrolling vertical column beside the detail panel instead of
stacking it above, so neither can cover the other; also collapse the
info block by default in 2/4-pane split view (toggle to expand) and
move the pane-count control into the detail header.
nntrivi2001