openSelector's onSelect fired the async handler with void and only
closed the overlay after a successful await. A rejecting handler (e.g.
setAgent -> setSession against a failing gateway) left the selector
stranded open with an unhandled rejection: the TUI froze on the picker
with no visible cause — a silent dead-end.
Root cause: failure path missing from the overlay lifecycle. The
handler now catches, surfaces the cause via chatLog, and always closes
the overlay.
Regression: /agent selection with a rejecting setSession asserts the
overlay closes and the cause reaches the chat log — fails pre-fix.
* fix(skills): expand explicit references on agent turns
Route generic Gateway, CLI, webhook, and local agent turns through the same explicit skill-reference renderer as channel auto-replies. Keep original transcript text, preserve unknown slash behavior, and fail visibly for allowlist-hidden skills.
Maintainer review: scoped Option 1 — generic agent turns expand both $skill-name and leading /skill-name args through shared skill rendering; they do not run the channel command dispatcher, and all other slash commands retain their existing behavior.
* fix(skills): bound explicit reference prompts
* fix(skills): prefer allowed reference collisions
* fix(skills): preserve command invocation boundaries
* fix(skills): reject hidden channel slash commands
* perf(skills): skip literal dollar discovery
* fix(ui): reveal clipped session history
Show a persistent earlier-history action for paginated sessions and reveal loaded content without chaining observer fetches.
Related: #110771
* refactor(ui): split chat history support
* fix: stop service child trees before timeout completion
* fix: preserve service child root results during cleanup
* test: isolate relay selection from linux spawn wrapping
* refactor: keep service anchor payload internal
* test: validate lifecycle probe pid output
* fix: close service child adversarial races
* fix: preserve service startup failures
* fix(process): flush service output before root result
* fix(process): bound cleanup on held output
* chore(process): satisfy promise executor lint
* fix(process): finish cleanup after lineage loss
* fix(process): retain relay output before subscription
* fix(ci): register service child runtime entries
* fix(ci): expose relay safety invariants
* fix(process): preserve service cancellation grace
* test(process): prove service grace on macOS
`checks-windows-node-test` ran the whole 222-266s Windows body in a single
Blacksmith lane, so every run that scheduled it finished at 287-334s against a
~180s plateau. Measured across 45 canonical main runs it was the wall in all
seven runs that triggered it (324-519s vs a 210s median elsewhere).
Blacksmith's Windows class admits exactly 2 concurrent jobs (run 31865243804),
which is why the earlier 3-part split failed: its tail queued 211s behind a
finished part. Pin the split width to that measured capacity instead. The two
parts are balanced by measured per-project wall time -- part 1 keeps the old
list 1 plus list 2's non-infra projects (~108s), part 2 takes list 2's infra
project plus the old list 3 (~112s) -- so both land near 113s and the lane
stops setting the wall.
One canonical partition now serves every backend; the github-mode/dispatch
3-part variant and its separate file inventory are gone, along with the
single-lane `task: test` case and the per-row `matrix.runner` the collapsed
lane needed. The complete 65-file Windows inventory is preserved exactly, and
the guards pin the width on both sides (mutation-checked: restoring 3 parts
fails the matrix guard).
ClawSweeper flagged that the prior matrix at line 31-48 tested
labels/title/notes by manually invoking normalizeBoundedString, even
though those fields use the independent normalizeTitle, normalizeNotes,
and normalizeLabels functions (each with its own generic message).
A regression that drops the (got N) suffix from normalizeBoundedString
would not be caught for those three surfaces — operators would still
see the old opaque error.
This commit narrows the matrix to only the consumers of the shared
helper (comment body, link title, link target, link URL, card id) and
restores the exact labels gateway assertion that should match the
unchanged normalizeLabels error.
Verification: 9/9 inline Node 18 AST harness (node /tmp/verify-pr-118888-v77.mjs);
parent-commit proof-gate (pre-PR helper without the got suffix) fails
4/9 of the post-fix expected messages.
Closes ClawSweeper review on PR #118888 (P2 + P3).
The shared normalizeBoundedString helper in extensions/workboard/src
threw an opaque error like 'comment body must be 2000 characters or
fewer.' whenever an oversize string hit any bounded field. Agents that
posted a long run summary or oversized link URL had no way to tell
how far over the limit they were, so they could not retry with a
trimmed payload.
Include the actual length in the error so the message tells the agent
exactly what to do next:
comment body must be 2000 characters or fewer (got 3502).
The format applies uniformly to every bounded field routed through
this helper (labels, link URL, link title, link target, comment
body, title, notes, worker log message, session key, run id,
protocol violation detail, etc.) since the helper is the single
canonical owner of the message.
Cover the new format with a focused unit test that pins the regex
for the comment-body repro from #118845 and exercises the six other
bounded fields listed in the issue. Update the existing
gateway.test.ts assertion to allow both the old and new format so
the format change is a forward-compatible improvement.
The Block Kit session card terminalized into a '🛠️ N tool calls · ⏱️ Ns'
footer, leaving a per-turn receipt in the Slack transcript on the default
progress path. PR #122976 deleted the same receipt from the native progress
surface but left this sibling rendering it.
Finished cards now keep only the durable diff stat and the session link;
tool-call and elapsed counters stay live working state. Also corrects the
documented channels.slack.streaming.mode default, stale since #122552.
The sidebar queued-message badge and the composer queue reused the clock
glyph that also marks attached automations, so a session with both showed
two identical clocks distinguishable only by tooltip. Add an outbox tray
icon (Lucide inbox with the roof swapped for a rising arrow) and use it
for queued-to-send surfaces; the clock stays automation/cron-only. The
mock dev fixture now flags one session with hasAutomation so the badge
renders in mocked-dashboard proofs.
* fix(delivery): unify terminal settlement ownership
Treat identityless adapter returns as potentially visible across channel, queue, and cron paths. Let recovery own terminal completion so ambiguity persists as notice debt instead of being double-settled or silently suppressed.
Co-authored-by: ruel225 <ruel225@users.noreply.github.com>
* refactor(delivery): narrow terminal internals
Remove now-unused internal exports after terminal-settlement ownership was consolidated.
* test(tts): preserve message runtime exports
Import and spread the actual message runtime so the focused mock retains every runtime binding while overriding only the durable send core.
---------
Co-authored-by: ruel225 <ruel225@users.noreply.github.com>
When a caller passed an explicit maxBytes with optimizeImages on (the
default), the source-read bound was inflated to max(maxBytes, 100MB
document cap). A channel with a 5MB attachment cap therefore let a
hostile or oversized URL buffer up to 100MB before the final size check
— 20x resource amplification, and the inflation applied before content
classification, so it was not limited to images.
The inflation exists so oversized-but-compressible originals can be
fetched and compressed under the delivery cap. Size that headroom off
the thing that justifies it: 4x the 6MB image cap (24MB), enough for
large phone photos, instead of the document cap.
Accepted tradeoff (named at the constant): originals above 24MB that
would have compressed under the cap now fail; the fetch error names the
bound. Callers without an explicit cap keep per-kind defaults.
Live proof on a real localhost HTTP server (no mocked fetch): an 8.1MB
PNG under a 5MB cap compresses to 3.75MB and delivers; a 30MB-declared
image is rejected at the Content-Length precheck in 7ms with 'content
length 31457280 exceeds maxBytes 25165824'.
Regression: headroom-bound test fails pre-fix (30MB passes the old
100MB bound); companion test proves compression headroom still works.
* fix(browser): support Chrome Web Store native bootstrap
* chore: keep browser release note in PR body
* docs(browser): document Store identity trust boundary
* docs(browser): correct Store recovery guidance
Worktree sessions now derive their worktree/branch name from the same
concise generated title the sidebar shows (bounded 8s wait, raw-prompt
slug then crustacean fallback), persisted once as displayName so the
chat-send pass never duplicates the model call. Dashboard title
scheduling moves from the dispatch chain's finally (after the whole
first turn) back to dispatch start, so long or interrupted first turns
no longer leave sessions silently untitled (regression from #122471).
slugifyWorktreeTitle truncates at word boundaries; dead
generatedDisplayName plumbing removed.
Peter Steinberger