Compound learning: the claude-code-cli-acp bridge returned 'Not logged in'
despite a working claude -p, due to (1) a too-thin spawn env (needs XDG_*/USER/
SHELL beyond HOME/PATH) and (2) macOS login-Keychain session isolation for
detached/headless processes. Six headless tasks misdiagnosed it as an upstream
gap. Cross-linked from the ACP runtime integration pattern doc.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- CodeRabbit: spinner class `spin` -> `animate-spin` (matches the card's other
Loader2 usages).
- CodeRabbit (major): tighten auth-failure detection so it only fires when the
WHOLE turn is the short "Not logged in" message (<=80 chars), not when a long
legitimate answer merely mentions the phrase — avoids false positives.
- CodeRabbit (major): expand the auth-signal test to assert the full invariant —
set on a not-logged-in turn, clear (unlink) on a real response, and NOT flag a
long answer that mentions the phrase.
pi-claude-cli acp-driver 5/5; typecheck clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Greptile P2: `acp.active` now reflects the ACTUAL dispatch determinant
(FUSION_CLAUDE_ACP, which includes the operator force-override), not the
experimental flag alone — so the status isn't misleading when forced on/off.
- CodeRabbit/Greptile P2: add FNXC:ClaudeAcp comments to the new code blocks
per the AGENTS.md greppable-comment convention.
Already fixed in the prior commit (daa37d08c): the P1 "sticky env" / latch
(applyClaudeAcpEnable now recomputes each call + FUSION_CLAUDE_ACP_FORCE
override) and the enable->disable-on-same-env regression test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When the bridged `claude` can't authenticate (detached daemon / no keychain),
the turn returns "Not logged in" instead of a real answer. Rather than silently
relay that, detect it and let the user choose.
- Driver: detect a "Not logged in"-only turn and write a cross-process signal
(fusion-acp-bridge-auth.json); a real response clears it (acp-driver test).
- Dashboard status: GET /providers/claude-cli/status reports
acp.authFailed + authReason from the signal.
- UI: the Claude CLI provider card shows an auth-failure banner with
"Use claude -p" (sets experimentalFeatures.claudeCliAcp=false) and
"I fixed auth — re-test", plus a fix hint (run `claude` to log in).
- Enable resolution now recomputes each call with an operator force-override
(FUSION_CLAUDE_ACP_FORCE), so the "Use -p" fallback takes effect on the next
turn — no restart. claude-acp-enable tests updated.
pi-claude-cli + engine tests green; dashboard typecheck clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
GET /providers/claude-cli/status now returns an `acp` block:
{ enabled (experimental flag), bridgeAvailable (KTD10 published path), active
(enabled && acpEnabled && bridgeAvailable) } so operators can see whether Claude
CLI is routing through the ACP bridge vs `claude -p` — important for the
default-on rollout. Additive; typechecks clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the manual FUSION_CLAUDE_ACP env enable with an experimental feature
switch. `experimentalFeatures.claudeCliAcp` is ON by default (off only when
explicitly set false); the engine translates it into the FUSION_CLAUDE_ACP
dispatch the pi-claude-cli provider reads, at registerExtensionProviders time.
- Still fail-closed: with no bridge path published (acp-runtime plugin absent),
the provider falls back to `claude -p`.
- Explicit FUSION_CLAUDE_ACP env always wins (operator / test override).
- New testable helper claude-acp-enable.ts (6/6 tests); flag documented in the
core experimentalFeatures doc.
So with the acp-runtime plugin installed, Claude CLI now routes through the ACP
bridge by default; set experimentalFeatures.claudeCliAcp=false to force `-p`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two-reviewer pass (security + architecture) on KTD10 + the full Route A
increment: no code defects, no P0, merge-ready as a dormant increment. Applying
the P1 follow-ups:
- Add the feature changeset (@runfusion/fusion minor) — the one convention gap.
- KTD10 tests: fail-closed (bridge not resolved -> env stays unset -> -p) and
idempotency (second onLoad keeps the first published path).
- Document the two intentional, parallel MCP-forwarding paths (U10 engine-adapter
vs U11 provider-driver) so nobody double-forwards, and the known
ACP-path-token-usage=0 residual so U12 doesn't treat it as a bug.
Reviewers confirmed: dormancy invariant holds end-to-end (nothing sets
FUSION_CLAUDE_ACP=1; both flag+path required; -p is the default); OAuth pi path
untouched. 206/206 plugin tests, 333/333 pi-claude-cli tests, typecheck clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The acp-runtime plugin's onLoad now publishes the identity-pinned bundled
claude-code-cli-acp path to FUSION_CLAUDE_ACP_BRIDGE (when unset), so the
pi-claude-cli kill-switch resolves the bridge WITHOUT a manual env var — no
engine->plugin static coupling. Publishes the path only; the ACP transport stays
OFF until an operator sets FUSION_CLAUDE_ACP=1 (rollout gate). Explicit env
override wins; resolver is pinned to the plugin's node_modules/.bin shim.
204/204 plugin tests green (3 new KTD10 tests); typecheck clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Live run: forwarded MCP tools and native Bash both refuse to execute when we
return cancelled to session/request_permission (no TOCTOU). streamViaAcp's
deny-by-default + break-early is verified safe. Env allow-list (incl. XDG/USER)
validated as required for the bridged claude to authenticate.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
streamSimple routes to streamViaAcp ONLY when FUSION_CLAUDE_ACP=1 AND a bridge
path is provided; otherwise stays on the -p streamViaCli path. Covers the three
cases (off / flag-without-path / flag+path) and asserts the bridge path + env
are forwarded. 333/333 pi-claude-cli tests green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three-reviewer pass (correctness/security/reliability) on the highest-risk file.
P0:
- Break-early now gates on isPiKnownClaudeTool: Claude's internal ToolSearch
(used to load deferred MCP tools) no longer aborts the turn before the real
fn_* call. Surface+break works from both tool_call updates and
request_permission. New test replays the U9 [ToolSearch, fn_task_list] sequence.
- Downgrade a tool_use turn that surfaced zero pi tool calls -> stop (mirrors
provider.ts), so pi never dispatches non-existent tools.
- register the bridge child in the process registry (no orphan on teardown).
- inactivity timeout (30 min, re-armed per chunk) + per-RPC timeouts on
newSession (a hung bridge now ends the stream and dies).
P1:
- capture bridge stderr + child 'close' handler -> surface exit code/stderr
(no more silent, undebuggable failures).
- sanitize untrusted agent output: strip ANSI/control chars, per-chunk +
per-turn caps, bound tool ids/names (no terminal-escape injection / DoS).
- validate bridge path (absolute + exists) before spawn.
- preserve image content blocks in the prompt (flatten-to-text dropped vision).
P2:
- enforce the bridge env allow-list INSIDE the driver (HOME/PATH/terminal only),
not trusting the caller-supplied object.
Documented residual (kill-switch stays OFF until verified): the bridge's
tool-execution ordering and native-tool (Bash/Read/Write) execution-prevention
need a live behavioral test before any lane enables this path.
pi-claude-cli: 330/330 tests green; typecheck clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds streamViaAcp: a drop-in alternative to streamViaCli that drives Claude
through the claude-code-cli-acp bridge over ACP instead of `claude -p`. Returns
the same AssistantMessageEventStream, so streamSimple dispatches to either
transport behind a kill-switch (FUSION_CLAUDE_ACP=1 + an injected bridge path),
OFF by default — the live `-p` path is byte-for-byte untouched until soak.
- Full-history prompt every turn (buildPrompt) — the ACP path has no --resume (R13).
- Forwards schema-only MCP servers so Claude emits correct tool calls; breaks
early on the first tool_call (cancel turn, surface to pi) so the bridge never
executes Fusion's tools — mirrors the `-p` break-early pattern.
- Translation reuses the tested createEventBridge by synthesizing Claude stream
events from ACP session/updates, sharing pi sequencing + tool-name mapping.
- Bridge env forwards only HOME/PATH so `claude` authenticates from the login
session (R17); never inherited process.env or API keys.
Verified: 3/3 translation unit tests; real-bridge session/update shapes confirmed
(agent_message_chunk text + tool_call); 326/326 existing pi-claude-cli tests green;
typecheck clean.
Remaining for Route A: engine injection of the bridge path (KTD10), U12 picker/
auth/status, U13 workflow verification.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reverse-engineered the streamViaCli contract and locked the U11 build plan:
AssistantMessageEventStream shape to match, the kill-switch branch point in
streamSimple, the KTD10 injection seam (pi-claude-cli adds @agentclientprotocol/sdk;
bridge path injected like mcpConfigPath), MCP-server construction reusing
ensureMcpConfig, full-history prompting (R13), ACP->pi event translation
paralleling event-bridge.ts, and live-bridge verification via the U9 harness.
No transport change yet — the live `claude -p` path is untouched.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Route A unblock + the first Route A code increment.
- U9 verdict recorded (plan OQ1 + docs/acp-contract.md): in an authenticated
interactive session the pinned claude-code-cli-acp 0.1.1 bridge forwards
session/new mcpServers to Claude, Claude invokes the forwarded Fusion tool,
and the call traverses the ACP permission gate (session/request_permission).
Both security-critical answers resolve positively — overturns the headless
NOT-GO chain (FN-6466/6467/6473/6476), whose only blocker was running
detached from the login keychain session.
- R17 (daemon auth) recorded and closed for the supported setup: creds are
macOS Keychain-only; the user's login-session fn daemon has keychain access
(the existing claude -p provider authenticates there), so the bridge does too.
- U10: thread an optional mcpServers list through the ACP runtime contract.
newAcpSession now forwards it (was hardcoded []); AgentRuntimeOptions (engine
+ plugin-local copy) gains the field; defaults to [] to preserve Route B's
read-only ask posture. Tool calls still route through the U5 permission floor.
Plugin typechecks clean; provider-session tests 12/12 (incl. 2 new forwarding
tests). U11-U13 (provider transport swap, picker/auth, workflow verify) remain.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Index the upstream ACP MCP forwarding sponsorship doc so the docs README and CLI index test keep it discoverable.
- Add the upstream ACP MCP passthrough and permission forwarding sponsorship doc link under Audit Reports.
- Add the upstream doc to the required docs README index coverage.
- Guard that CLI Printing Press docs remain indexed in Audit Reports only, not duplicated under Plugins.
Files changed:
docs/README.md | 5 +++++
.../cli/src/__tests__/docs-readme-index.test.ts | 21 +++++++++++++++++++++
2 files changed, 26 insertions(+)
Fusion-Task-Id: FN-6479
Fusion-Task-Lineage: f23afeba-a989-4552-8857-fe2984df6081
Document the authenticated ACP bridge rerun attempt and preserve the Route A blocked verdict.
- Update the ACP contract with FN-6476 readiness proof results showing the pinned bridge still reports an unauthenticated Claude session.
- Keep OQ1 answers unresolved because no forwarded Fusion tool invocation or permission-gate traversal was observed.
- Add FN-6476 status notes to the Claude ACP runtime plan so U9 remains NOT GO without a claude -p fallback.
Files changed:
docs/acp-contract.md | 11 ++++++++++-
docs/plans/2026-06-14-001-feat-claude-acp-runtime-plan.md | 3 +++
2 files changed, 13 insertions(+), 1 deletion(-)
Fusion-Task-Id: FN-6476
Fusion-Task-Lineage: f37dcc62-9758-47af-bb50-169e902211a5
Record the upstream sponsorship package for ACP MCP passthrough and permission gating while keeping Route A blocked.\n\n- Add a ready-to-file upstream issue for claude-code-cli-acp MCP passthrough and permission-forwarding support.\n- Link the filed upstream issue from the ACP contract and runtime plan.\n- Preserve the OQ1/U9 NOT GO status until authenticated reruns prove forwarded tool invocation and gating.\n\nFiles changed:\n docs/acp-contract.md | 5 +\n .../2026-06-14-001-feat-claude-acp-runtime-plan.md | 3 +\n ...laude-code-cli-acp-mcp-permission-forwarding.md | 121 +++++++++++++++++++++\n 3 files changed, 129 insertions(+)
Fusion-Task-Id: FN-6475
Fusion-Task-Lineage: 3cbd14e3-9388-4cf4-b256-3678459eb926
Record the authenticated Route A ACP escalation rerun and keep U9 blocked.
- Update the ACP contract OQ1 status to FN-6473 with the observed unauthenticated bridge result.
- Capture the explicit request-permission instrumentation outcome: zero forwarded tool calls and zero permission callbacks.
- Extend the Claude ACP runtime plan with the escalation status and upstream bridge/ACP permission-forwarding requirement.
Files changed:
docs/acp-contract.md | 13 +++++++++++--
docs/plans/2026-06-14-001-feat-claude-acp-runtime-plan.md | 3 +++
2 files changed, 14 insertions(+), 2 deletions(-)
Fusion-Task-Id: FN-6473
Fusion-Task-Lineage: 2e6ec23b-a9cd-4fd0-b24e-72ee8bb145ac
Record the FN-6467 U9 rerun outcome so Route A remains gated on authenticated ACP bridge evidence.
- Document the second direct bridge attempt against claude 2.1.177 and claude-code-cli-acp 0.1.1.
- Preserve the accepted 62-tool custom-tools MCP payload details and lockfile integrity evidence.
- Mark forwarded tool invocation and ACP permission-gate traversal as unresolved because the bridge still reports Not logged in.
Files changed:
docs/acp-contract.md | 15 ++++++++++-----
docs/plans/2026-06-14-001-feat-claude-acp-runtime-plan.md | 3 +++
2 files changed, 13 insertions(+), 5 deletions(-)
Fusion-Task-Id: FN-6467
Fusion-Task-Lineage: fa2c1e16-7822-421b-80cf-e4ab247641a5
Record the real Route A U9 bridge attempt and keep OQ1 at NOT GO.
- Document that claude-code-cli-acp accepted a non-empty Fusion MCP server declaration.
- Capture the unauthenticated claude blocker before forwarded tool invocation or permission telemetry.
- Keep FN-6460 blocked until an authenticated rerun proves tool forwarding and ACP permission traversal.
Files changed:
docs/acp-contract.md | 13 +++++++++++--
docs/plans/2026-06-14-001-feat-claude-acp-runtime-plan.md | 3 +++
2 files changed, 14 insertions(+), 2 deletions(-)
Fusion-Task-Id: FN-6466
Fusion-Task-Lineage: 0bce3b9f-fbab-4562-9a7b-9d5551335fa4
Record the unrecovered Route A U9 verdict and the design-only U14 confirmation so follow-up work preserves the ACP permission-gate blocker.\n\n- Add an ACP contract open-question entry that marks Route A as NOT GO until real MCP forwarding and ACP permission-gate traversal are proven.\n- Capture the FN-6465 recovery outcome in the Claude ACP runtime plan, including the missing FN-6459 evidence and required rerun criteria.\n- Confirm the internal Route A design mechanisms while keeping implementation blocked on U9.\n\nFiles changed:\n docs/acp-contract.md | 24 ++++++++++++++++++++++\n .../2026-06-14-001-feat-claude-acp-runtime-plan.md | 9 ++++++++\n 2 files changed, 33 insertions(+)
Fusion-Task-Id: FN-6465
Fusion-Task-Lineage: d4762440-cbb9-4206-b4b2-be578eba30dd
Ensure AI-merge clean-room directories are cleaned up even when setup fails before worktree registration.
- Move clean-room directory creation inside the cleanup guard so terminal setup failures can remove it.
- Skip cleanup only when no merge root was ever created, while still unregistering any registered paths.
- Add regression coverage for active-session registration failure after mkdtemp and before git worktree add.
Files changed:
.../ai-merge-worktree-cleanup.test.ts | 31 ++++++++++++++++++++++
packages/engine/src/merger-ai.ts | 21 ++++++++++-----
2 files changed, 45 insertions(+), 7 deletions(-)
Fusion-Task-Id: FN-6453
Fusion-Task-Lineage: 364b329b-c3e7-4902-8ca8-673360b0d9ac
The non-blocking Full Suite tier on main was red on shards 2 and 4:
- roadmap-store schema assertion lagged core's SCHEMA_VERSION bump to 117
(landed in FN-6277), so it still expected 116.
- useCeSessions "cancel surfaces a transport error" failed deterministically:
a session with an in-flight status keeps the poll fallback running, and a
successful background list refresh called setError(undefined), wiping the
cancel error before it could be observed. Background refreshes (poll + push)
now leave action errors intact; only user-initiated/initial refreshes clear.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Document the restored FN-6437 proof-point results in the external plugin runbook.
- Replace the pending-verification placeholder with the recorded NOT MET verdict for @runfusion/fusion@0.43.0.
- Add package, integrity, environment, command, scaffold, and TypeScript failure evidence from the restored report.
- Preserve explicit gaps for list/enable proof and follow-up checks that were blocked by the released scaffold failure.
Files changed:
docs/plugins/external-proof-point-runbook.md | 132 +++++++++++++++++++++++++--
1 file changed, 124 insertions(+), 8 deletions(-)
Fusion-Task-Id: FN-6452
Fusion-Task-Lineage: 14d3b848-59cd-43bb-b725-6586182906ee
Add a durable proof-point runbook for validating external plugins against released Fusion CLI builds.
- Document release selection, npm integrity capture, scaffold/build/test/install/enable validation, and pass/fail criteria for external plugin proof points.
- Link the runbook from the docs index so it remains discoverable with other plugin authoring docs.
- Extend the docs README index test to require the new proof-point runbook entry.
Files changed:
docs/README.md | 1 +
docs/plugins/external-proof-point-runbook.md | 211 +++++++++++++++++++++
.../cli/src/__tests__/docs-readme-index.test.ts | 1 +
3 files changed, 213 insertions(+)
Fusion-Task-Id: FN-6438
Fusion-Task-Lineage: 5c768d7e-4465-42ac-9535-f16dc42ab6d0
Enable mobile users to swipe across the agent detail tab strip.
- Restores horizontal touch panning on the agent detail tabs while preserving vertical pan behavior.
- Documents the mobile touch-action requirement next to the tab-strip CSS.
- Adds a mobile regression test covering pan-x touch-action and horizontal overflow.
Files changed:
packages/dashboard/app/components/AgentDetailView.css | 6 ++++++
.../__tests__/AgentDetailView.mobile-scroll.test.tsx | 15 +++++++++++++++
2 files changed, 21 insertions(+)
Fusion-Task-Id: FN-6450
Fusion-Task-Lineage: 7006af83-4490-4fe0-a357-7d19b7b522c0
Rescue the session reconnect SSE test by shutting down route-owned background work before temp cleanup.
- Isolate the focused API harness from TaskStore EventEmitter startup workers.
- Dispose API routes and scheduled AI session cleanup before deleting the test temp root.
- Remove session-reconnect from the dashboard quarantine ledger and vitest skip list.
Files changed:
.../src/__tests__/session-reconnect.test.ts | 22 ++++++++++++++++++----
packages/dashboard/vitest.config.ts | 6 ++++--
scripts/lib/test-quarantine.json | 5 -----
3 files changed, 22 insertions(+), 11 deletions(-)
Fusion-Task-Id: FN-6447
Fusion-Task-Lineage: a3fef28c-b92b-4f47-9181-a63bb31e4f48
Ensure the Secrets page root expands within the project content flex row on narrow screens.
- Let the Secrets view root flex to available width with a zero min-width.
- Preserve existing height and spacing behavior while avoiding intrinsic card-width collapse.
- Cover the mobile layout contract with a root-container CSS assertion.
Files changed:
packages/dashboard/app/components/SecretsView.css | 7 +++++++
.../dashboard/app/components/__tests__/SecretsView.mobile.test.tsx | 7 +++++++
2 files changed, 14 insertions(+)
Fusion-Task-Id: FN-6446
Fusion-Task-Lineage: 34f55d8e-ee57-432c-b1ce-165cc43333ee
