tools/ui/build.sh and serve.sh still cd'd into the app/ directory the
flattening removed, so make test-e2e / ui-dev / ui-smoke died at the
first line. The staged Gitea workflow had the same stale cd in every
job, plus a coverage gate with no coverage run before it — it now
goes through the make targets (tooling discipline: the make layer owns
env setup) with make test-coverage feeding make coverage-gate.
Fixing the paths exposed the real break: flutter build web --wasm
cannot compile the tree since the dart:ffi pivot (tree-sitter, native
PTY) — dart:ffi does not exist on the wasm target. Fence vs park vs
drop is filed as Q-50; the workflow's e2e job is withheld with a
pointer there, and T-384 sits in review until Q-50 resolves.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New governance/questions/design.md holds one open question per Part IV
feature proposal — Tier 1 (agent blame, context x-ray, trust ledger,
activity HUD, active-ticket chip), Tier 2 (twin-timeline rewind, visual
dialog, immortal terminals, cost ledger, ticket dispatch), Tier 3
(semantic terminal, codebase map, live mixed documents, sealed
workspace) — plus one batch record for the honorable mentions, so each
can resolve into a D-record + initiative or an R-record. Remote Claude
over SSH got a context append on existing Q-23 instead of a duplicate
record. README index regenerated by pql decisions sync.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The detection/follow core for the live-tail sub-card, with the UI wiring
to follow. Claude Code runs every Bash tool itself and clide only sees the
final tool_result block — we can't mirror the running process, so instead
we detect a file-backed source the command follows and open our own
read-only follower on the same file.
- bash_tail_source.dart: detectBashTailSource() parses a Bash command for a
single, safe, file-backed source (tail/cat/less with one file arg, inside
the workspace via resolveUnderRoot). Returns null for a pipe-into-tail, a
redirect, two files, or a path outside the repo — the caller then shows a
"nothing to follow" note. bashHasTailIntent() gates WHEN the segment
appears: v1 triggers on `tail`/follow-flags only, so ordinary cat/ls/git
cards stay clean (cat/less remain detectable for later).
- file_tail_follower.dart: a polling, read-only `tail -f`-style follower
(no subprocess, no touching Claude's command) that emits the trailing
window then appended deltas, and re-reads from the top on truncation.
Tested: 19 parser cases (incl. the `git push | tail -25` and outside-
workspace null cases), the intent predicate, and the follower (initial
window / appended delta / missing file / rotation / start / stop).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Records the T-7 -> T-317 merge on D-91's Relationship bullet (the
former Tier-5 canvas+graph epic folds into the unified canvas
epic; canvas pane T-322 + graph view T-323 consume the shared
renderer) and the D-91 index line. Plus the changelog for this
turn's board restructure: epic T-317, core engine T-318,
templates T-319/320/321, re-parented T-313/315/316, T-7 cancelled
as superseded, T-41 moved to T-276.
Governance + .pql/changelog only — leaves concurrent lib/ work
and CHANGELOG.md to the other worker on this branch.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A clide:// link is an untrusted external vector (any webpage can fire one), so it
no longer translates to a command in parseArgv. It routes the raw URL to a new
builtin.deeplink handler that is doubly defensive: a default-deny allowlist
(kDeepLinkSafeActions — only the read-only 'open' verb; run/git/write/passthrough
rejected) AND a mandatory 'an external link wants to: … allow?' confirmation
before anything runs. Records the security boundary as D-90.
The earlier silent editor.open passthrough is replaced; open still works, now
behind the prompt. Tests cover the allowlist (the boundary) + the gating.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
T-235: app-scoped kActivityFoldLevelKey + a claude.activity.fold-level command
that cycles none→tools→thinking→everything; ClaudePane + team_panel_host read it
and re-fold live via the settings notifier. Unit tests for the helpers.
T-132 cleanup: the one blocked item (account/team token budget) is detached
(T-158), reframed as Q-34 'how + when to surface the budget given upstream
doesn't expose it', with T-158 as its backlog resolver. T-132 closed — all
doable work delivered.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A pasted-image @<path> token now renders as an inline, keyboard-activatable
thumbnail in the Claude conversation that opens the full image in the lightbox;
the composer's attachment chips use the same (larger, 44px) thumbnail. New
ImageThumbnail + openImageLightbox in the Claude layer; ClideMarkdown gains an
onImageToken builder seam (mirroring onRecordTap) that drops a WidgetSpan into
the text flow — it owns no Image.file/lightbox, staying generic. Missing files
degrade to a placeholder; render-only (the sent text + copyText are unchanged).
Resolves the conflicting T-236 (inline thumbnail) / T-254 (image card) designs
into the hybrid the user chose; recorded as D-89.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Nine surfaces hand-rolled the same anchored-overlay + row-list + barrier +
keyboard-nav pattern. Extract one owned primitive (no Material):
- ClideAnchoredOverlay (clide_anchored.dart): positioning + lifecycle —
LayerLink/CompositedTransformFollower or centred Positioned, side/align +
auto-flip on viewport bounds, full-screen tap-away barrier, OverlayEntry
bookkeeping, focus capture, Esc-to-close. Driven by a ClideOverlayController.
- ClideMenu + ClideMenuListController (clide_menu.dart): a dropdown-token row
surface (items + separators) with arrow/enter/escape nav, skip-disabled,
active mark, per-item colour/leading glyph, keepOpenOnSelect (live-apply),
and onArrowLeft/Right hooks. The nav controller is reusable by surfaces that
keep bespoke rows (typeaheads, quick-open).
Additive — no call sites changed yet. D-88 records the convention (new `design`
domain): anchored pickers build on these; modal pickers stay on DialogRouter.
Tests: clide_anchored_test (open/close, barrier, Esc, centred, clean dispose)
and clide_menu_test (list-nav skip/wrap, select + onClose, disabled, Esc,
keepOpenOnSelect; pure ClideMenuListController cases).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
D-87 — T-54's "output and log panel" is a bottom output dock: read-only,
two tabs (Output = the Logger stream, filterable + auto-scroll; Problems =
diagnostics moved out of the sidebar). Toggled by a single status-bar widget
that replaces the app-status indicator (merged health+log: green check when
clean, warn/error counts otherwise) via click or Cmd/Ctrl+J. Needs a bounded
ring sink on the Logger (no history today). Amends D-47: the dock is the one
surface allowed to push Claude up, capped so Claude stays >=50%.
Resolves Q-28 by splitting on interaction: read-only output (logs/problems)
goes in the dock; the terminal does NOT — it stays a first-class editor-pane
surface, tracked by new T-258 (swap-vs-split, with Q-27). Refines T-54.
Wireframe under docs/design/wireframes/output-dock/.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
D-85 — event bus delivery semantics, resolving Q-2 + Q-3: bounded
per-subscriber ring buffer with drop-oldest back-pressure (producer never
blocks, subscribers never killed) and a gap marker so a pull reader detects
loss; in-memory cursor retention serves `clide events --since`; if
persistence is ever needed it is bus-owned, not a subscriber-subsystem —
reversing ADR 0006's unreasoned open-questions footer.
D-86 — MCP tool surface, resolving Q-32: expose the full mcp__clide__*
namespace but generate tools/list from the co-registered command registry
(D-74) that already feeds CLI + palette, so breadth costs no second
maintained surface; per-command MCP opt-out for poor-fit verbs.
Q-33 drift-fixed: was already closed by D-73 (SSE only) but the index still
listed it open; re-confirmed SSE-only, stdio/WS not added.
Refines T-223 (cursor pull events) and T-225 (MCP reachable) — both now
unblocked and scoped. Index regenerated by pql decisions sync.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Records D-83 answering Gap 5 of docs/self-analysis.md: clide commits to
both "Claude inside clide" models, with the clide-HOSTED stream-json
session (D-77/D-78) as the PRIMARY dogfood target — the process clide
spawns, hence the one Epic B (T-214) bootstraps (CLIDE_SOCK/CLIDE_WORKSPACE
+ PATH + context note + Bash(clide *) allow rule). The EXTERNAL CLI driver
(D-68) stays a first-class but SECONDARY, best-effort integration: manual
install (T-212), no promise to observe its non-clide tool use.
Scopes the D-6 parity contract (Epic C / T-218) to clide's own surfaces
reflected through the CLI in both directions; an external agent's
side-channel reads/tests/git are explicitly out of parity scope.
Closes T-224 (under T-208 "Give Claude hands").
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Documents the decision behind T-205: keymap sequences are space-
separated (list still alternates, + still joins a chord), repeat counts
apply by firing N times, and sequence matching is a reusable stateless
query + stateful matcher consumed at the interception point — because
the global KeyboardListener can't swallow events, so the editor's
Focus.onKeyEvent (T-206) is where normal-mode keys get consumed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Documents why the markdown/decisions readers load from a retained
per-reader nav-history (grab-current-on-mount + single 'load' path)
rather than per-view state (dies with the widget — the T-196 bug) or
MessageBus retention (wrong layer). The user chose the nav-history
helper over a bus fix.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The reader opened repo-local .claude markdown but rejected user-scope
files under ~/.claude with "path outside workspace" — that dir is
global, outside the repo, and files.read was repo-confined (T-102).
Per D-76 the Claude config surface is clide-managed, so files.read now
resolves a path under an allow-list: the workspace root plus trusted
extra read roots (FilesService.extraReadRoots), wired in main.dart to
~/.claude when present. Reads widen; writes stay repo-confined, and the
symlink re-check still refuses a config-root symlink that escapes. Off-
root paths and `..` traversal are rejected as before.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Find-in-files / replace (T-52/T-53) run as an in-process isolate-pool
grep engine behind an engine-agnostic search.grep verb — not pql (its
search is a ranked document index, with no line numbers, regex, or
glob) and not a ripgrep shell-out (unvendored, not guaranteed
cross-platform). ripgrep is kept as a future optional accelerator
behind the same verb. Clarifies the D-3 wrap-pql boundary: content
grep is a code-navigation primitive pql does not offer.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Regenerated by `pql decisions sync` — moves resolved questions (Q-6, Q-19,
Q-21, Q-22) into a Resolved section.
Co-Authored-By: Claude <noreply@anthropic.com>
Empirical spike against claude 2.1.150 (driving the real CLI over
stdin/stdout + reading the shipped binary's zod schemas) pinned the
wire shapes for the stream-json control protocol: the can_use_tool
permission request, the control_response envelope, the
--permission-prompt-tool stdio enabler (without it "ask" tools silently
auto-deny), the allow-requires-updatedInput quirk, and AskUserQuestion
answered via updatedInput.answers. Captured in a version-pinned spike
note with a resilience section (detection canaries + a ranked fallback
menu) so a future Anthropic change to this undocumented contract doesn't
leave us at a blank slate.
D-78 records the decision: permissions ride the stdio control channel,
not MCP (MCP is reserved for capability/tool provision); refines D-77.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
test / unit + widget + golden + a11y (push) Failing after 29s
test / integration_test (xvfb) (push) Has been skipped
test / bundle smoke (xvfb 5s) (push) Has been skipped
test / daemon subprocess + web WASM smoke (push) Has been skipped
test / dart doc (lib API) (push) Failing after 27s
Accepted (phased). Pivot the Claude pane from the interactive tmux TUI to
the stream-json control protocol: structured events instead of transcript
tailing, permissions + AskUserQuestion handled natively via canUseTool,
persistence via --resume. Claude's tmux agent-team mode is headless-
incompatible, so teams become clide-orchestrated — N managed sessions
coordinated by a clide-hosted MCP broker, with team-awareness injected via
--append-system-prompt/--agents.
Captures the unified-session-model upside: teammate / secondary tab /
forked branch / inline subagent collapse into one primitive (a managed
session rendered as a pane), with the sidebar as the cockpit. Amends D-41
(persistence) and evolves D-75 (rendering source). Phase 1 single-agent
first; phase 2 the unified model.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Extends D-75's accepted CC-internals coupling from the transcript/team
schema to the config layout: a builtin-owned ClaudeConfig service is the
app-wide source of truth for skills, commands, settings, and permissions
(global + local, layered), watched and refreshable. Built-in slash
commands come from a stream-json probe cached per claude version id.
Kernel stays Claude-agnostic — Claude is a non-disableable extension but
still an extension. Implemented by T-151..T-154.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Decision for epic T-132: clide renders the Claude conversation as native
Flutter widgets driven by Claude Code's transcript JSONL, not by scraping
the PTY/TUI. Claude still runs under tmux for process persistence (D-41),
but its content is sourced from the transcript. The terminal emulator is
retained as a general IDE tool only. Records the Claude-centric > CLI-first
ordering (D-6 surfaces preserved) and the accepted, version-pinned coupling
to Claude Code internal contracts, isolated behind one reader/observer.
Co-Authored-By: Claude <noreply@anthropic.com>
Captures the T-120 design decision: the typed argument schema travels
with each command (built-in modules and extension contributions both
register their own) rather than living in a central static map, which
would be blind to extension-contributed commands and fight the plugin
model (D-46). The dispatcher accumulates a cmd->schema registry and
validates req.args before invoking the handler. The T-104 git
validators stay as defense-in-depth because the git client is also
reachable directly from the UI. MCP tools/list generation deferred to
the T-130 track.
Co-Authored-By: Claude <noreply@anthropic.com>
test / unit + widget + golden + a11y (push) Failing after 27s
test / integration_test (xvfb) (push) Has been skipped
test / bundle smoke (xvfb 5s) (push) Has been skipped
test / daemon subprocess + web WASM smoke (push) Has been skipped
test / dart doc (lib API) (push) Failing after 28s
D-56 and D-68 gain dated implementation-notes amendments pointing at
the eight T-99 slices (T-124 server through T-131 wrap-up) and the
D-70/71/72/73 records the work spawned. D-68's main Decision text
swaps "Transport deferred to Q-33" for the closed reference to D-73.
CONTRIBUTING.md gains a "Running clide from the shell" section
covering the verb surface, exit-code contract per D-68, and
`/ide` MCP discovery via `~/.claude/ide/<pid>.lock`. governance/
README.md picks up the D-73 entry it was missing.
T-119 moves out of backlog (its IPC dispatch path now exists);
T-120 gains a 2026-05-19 re-scope note pointing the typed schema
framework at the now-real argv grammar / wire contract. T-99
itself transitions to done.
Co-Authored-By: Claude <noreply@anthropic.com>
test / unit + widget + golden + a11y (push) Failing after 26s
test / integration_test (xvfb) (push) Has been skipped
test / bundle smoke (xvfb 5s) (push) Has been skipped
test / daemon subprocess + web WASM smoke (push) Has been skipped
test / dart doc (lib API) (push) Failing after 1m0s
Three architectural commitments T-124 needs settled before any code:
* D-70 — socket path is per-workspace and deterministic
(`$XDG_RUNTIME_DIR/clide/<sha256(workspace)[:16]>.sock`). No env
override; matches the per-repo tmux socket pattern (D-41) so users
see one consistent shape across clide's process boundaries.
* D-71 — file permissions are the only gate: `0600` on the socket,
`0700` on the parent. Capability-token auth stays a Tier-6 concern
tracked by Q-1.
* D-72 — multi-connection accept loop, serial dispatch on the main
Flutter isolate. Per-handler isolate offload (the NativePty /
SchedulerService pattern) is available where it matters; the IPC
layer doesn't impose isolate-safety across every subsystem.
T-124's description updated to inline the three decisions so the
implementer starts with the contract, not a discovery pass.
Co-Authored-By: Claude <noreply@anthropic.com>
The expanded canonicalPairs from T-114 (muted text, status chips,
syntax tokens on the code-block surface, panel focus border) made the
four named themes fail WCAG-AA. Retuning their palettes to pass would
have changed the look users picked them for, so the gate is split
instead.
`canonicalPairs` shrinks back to the baseline every named theme passes;
the new `extendedPairs` carries the stricter set and only runs against
themes whose name ends `-hc` or `-cb`. Sibling files (`clide-hc`,
`midnight-hc`, `paper-hc`, `terminal-hc`) ship today; the policy lives
in D-69 with a back-ref from D-22.
Co-Authored-By: Claude <noreply@anthropic.com>
D-68 (dual integration surface — Bash CLI primary, MCP secondary)
landed in 8074bf4 inadvertently; this lands its index pointer in
README.md and the two open questions D-68 spawned.
- Q-32: tool surface — minimum /ide vs extended `mcp__clide__*`.
- Q-33: transport — SSE / WebSocket / stdio / all.
No behavior change. Governance bookkeeping only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
test / unit + widget + golden + a11y (push) Failing after 35s
test / integration_test (xvfb) (push) Has been skipped
test / bundle smoke (xvfb 5s) (push) Has been skipped
test / daemon subprocess + web WASM smoke (push) Has been skipped
test / dart doc (lib API) (push) Failing after 1m3s
Two open questions have been answered by reality and should be
formally resolved:
- Q-21 (pql planning home): pql 1.4.30 ships the full surface and
clide consumes it via shell-out. Resolves to D-3 + D-39, which
already encoded the intent.
- Q-22 (ticket persistence): pql 1.4.x's append-only per-month SQL
changelog with inline LWW guards is option (3) of Q-22 evolved
into a form that merges by default. Clide migrated to it on
2026-05-09.
Adds D-67 to formalise clide's adoption: which directories carry
changelog data, what stays gitignored, which hooks run when. Q-22
points at D-67; Q-21 points at the pre-existing D-3 + D-39.
Pql state: 106 decisions → 107 (D-67 added); 134 refs → 139; 0
broken; validate ok.
Co-Authored-By: Claude <noreply@anthropic.com>
test / unit + widget + golden + a11y (push) Failing after 31s
test / integration_test (xvfb) (push) Has been skipped
test / bundle smoke (xvfb 5s) (push) Has been skipped
test / daemon subprocess + web WASM smoke (push) Has been skipped
test / dart doc (lib API) (push) Failing after 1m2s
pql 1.4.30 introduces a per-type subdirectory layout for decision
records: governance/{decisions,questions,rejected}/<domain>.md. The
parser infers record type from the parent subdir and domain from
the filename stem, replacing the flat decisions/<file>.md layout
with questions-X and rejected prefixes.
Moves:
- 6 D-files to governance/decisions/ (accessibility, architecture,
extensions, process, testing, tooling)
- 5 Q-files to governance/questions/ (stripped the questions- prefix)
- rejected.md split by domain to governance/rejected/{architecture
(5), process (4), accessibility (1), testing (1)}.md — split per
cross-reference target.
Drops:
- decisions/README.md (auto-regenerated as governance/README.md by
pql init below the records marker)
- decisions/questions.md (hand-curated index pointing at the per-
domain files; redundant once pql generates the index, and the
"load-bearing" curation was already present in each Q-record's
Context field).
Convention-driven; not a clide-specific change. Counts unchanged
across migration: 106 decisions, 95 tickets, 134 refs, 0 broken.
Co-Authored-By: Claude <noreply@anthropic.com>
