fix(agent): launch agent uv-run subprocesses with --no-sync

Agents with a write workspace (developer/product_owner/head_marketing/documenter)
run with cwd = their git workspace clone. Claude Code launches each MCP server
(flow/do/git-readonly/optimal/docs/search) and the SDK server as
`uv run python -m ...` from that cwd. When the clone's uv.lock drifts from the
baked image, `uv run` re-resolves and re-syncs /app/.venv against the clone's
lock — a multi-minute stall on a cold wheel cache — so the servers never reach
"connected": they sit at status="pending" and the agent gets ZERO gateway
verbs. It then can't claim/commit/idle (all MCP verbs), its Stop is rejected,
and it respawns in a loop redoing work it can't submit.

UV_PROJECT_ENVIRONMENT pins the venv location but does NOT stop the cwd-relative
resolve/sync (confirmed empirically on uv 0.11.1); `--no-sync` does, so the
servers reuse the baked /app/.venv as-is and start instantly. The /app-cwd roles
(qa/cell_pm/main_pm/auditor) were unaffected because their env already matches.

- orchestrator.py: --no-sync on all 6 generated MCP servers
- docker/scripts/sdk-startup-hook.sh: --no-sync on the agent_sdk.server launch
- test_spawn_strict_mcp.py: assert every server's args start with run,--no-sync

docker/scripts/sdk-startup-hook.sh

@@ -18,15 +18,18 @@ PRECOMPACT_FILE="/tmp/roboco-precompact-${AGENT_ID}.md"
 # set (torch/lancedb/pyarrow/scipy, ~350MB) into a fresh venv. On a cold
 # uv wheel cache (first spawn after an image rebuild) that download takes
 # minutes and the SDK/MCP layer never comes up before the agent reaps.
-# Pin uv to the pre-baked image venv so it starts instantly regardless
+# Pin uv to the pre-baked image venv AND pass `--no-sync` on the run below.
-# of cwd. (The orchestrator sets the same var in every MCP server's env
+# Pinning the env location alone is NOT sufficient: `uv run` still discovers
-# in the generated mcp-config.json — keep both in sync.)
+# the cwd project and re-syncs the pinned venv against the clone's (drifted)
+# lock — that resync is the actual multi-minute stall. `--no-sync` skips it.
+# (The orchestrator passes the same var + --no-sync to every MCP server in
+# the generated mcp-config.json — keep both in sync.)
 export UV_PROJECT_ENVIRONMENT=/app/.venv
 
 # --- SDK bring-up ---------------------------------------------------------
 if ! curl -sf "http://localhost:${SDK_PORT}/health" >/dev/null 2>&1; then
     echo "[SDK] Starting for agent ${AGENT_ID} on port ${SDK_PORT}..."
-    nohup uv run python -m roboco.agent_sdk.server > "$LOG_FILE" 2>&1 &
+    nohup uv run --no-sync python -m roboco.agent_sdk.server > "$LOG_FILE" 2>&1 &
     SDK_PID=$!
     sleep 2
     if curl -sf "http://localhost:${SDK_PORT}/health" >/dev/null 2>&1; then

roboco/runtime/orchestrator.py

@@ -2015,8 +2015,12 @@ class AgentOrchestrator:
             # but on a cold cache (first spawn after an image rebuild) the
             # download takes minutes and the MCP servers never come up
             # before the agent burns its budget. Pinning the project env
-            # to the pre-baked venv makes `uv run` reuse it instantly,
+            # to the pre-baked venv is necessary but NOT sufficient: `uv run`
-            # regardless of cwd.
+            # still resolves the project from the workspace cwd and re-syncs
+            # when the clone's uv.lock drifts from the image — leaving the MCP
+            # servers stuck at status="pending" so the agent gets zero gateway
+            # verbs. Each server is therefore launched with `uv run --no-sync`
+            # (below) to use /app/.venv as-is and start instantly.
             "UV_PROJECT_ENVIRONMENT": "/app/.venv",
         }
 
@@ -2031,19 +2035,19 @@ class AgentOrchestrator:
             # Intent verbs — every role-scoped lifecycle transition.
             "roboco-flow": {
                 "command": "uv",
-                "args": ["run", "python", "-m", "roboco.mcp.flow_server"],
+                "args": ["run", "--no-sync", "python", "-m", "roboco.mcp.flow_server"],
                 "env": mcp_env,
             },
             # Content tools — commit, push, PR, journal, notify, message.
             "roboco-do": {
                 "command": "uv",
-                "args": ["run", "python", "-m", "roboco.mcp.do_server"],
+                "args": ["run", "--no-sync", "python", "-m", "roboco.mcp.do_server"],
                 "env": mcp_env,
             },
             # Read-only git views — status, log, diff, branches.
             "roboco-git-readonly": {
                 "command": "uv",
-                "args": ["run", "python", "-m", "roboco.mcp.git_readonly"],
+                "args": ["run", "--no-sync", "python", "-m", "roboco.mcp.git_readonly"],
                 "env": mcp_env,
             },
             # Knowledge base — RAG / semantic search / ask_mentor.
@@ -2051,6 +2055,7 @@ class AgentOrchestrator:
                 "command": "uv",
                 "args": [
                     "run",
+                    "--no-sync",
                     "python",
                     "-m",
                     "roboco.mcp.optimal_server",
@@ -2075,6 +2080,7 @@ class AgentOrchestrator:
                 "command": "uv",
                 "args": [
                     "run",
+                    "--no-sync",
                     "python",
                     "-m",
                     "roboco.mcp.docs_server",
@@ -2097,6 +2103,7 @@ class AgentOrchestrator:
                 "command": "uv",
                 "args": [
                     "run",
+                    "--no-sync",
                     "python",
                     "-m",
                     "roboco.mcp.search_server",

tests/unit/runtime/test_spawn_strict_mcp.py

@@ -92,3 +92,13 @@ class TestMcpConfigPinsBakedVenv:
                 f"/app/.venv — without it `uv run` re-downloads deps into a "
                 f"cwd-relative venv on every spawn (#179). env={spec['env']}"
             )
+            # Pinning the env location is necessary but NOT sufficient: from a
+            # workspace-clone cwd `uv run` still discovers the clone project and
+            # re-syncs the pinned venv against its drifted lock, which stalls and
+            # leaves the server stuck at status="pending" (zero gateway verbs).
+            # `--no-sync` skips that resync — it must come right after `run`.
+            assert spec["args"][:2] == ["run", "--no-sync"], (
+                f"MCP server {name!r} must launch with `uv run --no-sync ...` so "
+                f"a drifted workspace-clone lock can't trigger a resync stall; "
+                f"got args={spec['args']}"
+            )