✔ sessionModel: agent vars win, then the system's execution settings (8.970286ms)
✔ a pi bundle: prompt, policy, tools and manifest, 0600 in a 0700 directory (4.238711ms)
✔ a claude-code bundle adds the wrapped gate hook and the MCP config (2.846735ms)
✔ the prompt's built-in tools line is the harness's own names, as the manifest lists them (11.702305ms)
✔ the prompt names the project clones in the workspace, and none when there are none (7.778364ms)
✔ the prompt explains the review loop to a role with push_working, and reviews to one with fetch_review (8.922697ms)
✔ a bundle is written once: an existing file refuses (1.538376ms)
✔ a path with a single quote can't go into the hook command (1.477896ms)
✔ allow exits 0, a deny exits 2 with the reason on stderr (113.868933ms)
✔ a missing or wrong policy, or a bad event, exits 2 (83.573944ms)
✔ the bundle's wrapped command: a missing gate or node still blocks (1087.259508ms)
✔ claude: typed tools through MCP, the hook blocks, builtins outside --tools don't exist (769.143007ms)
✔ claude: the hook alone blocks a path outside the workspace (449.357668ms)
✔ claude: a second turn resumes the first turn's session (769.594822ms)
✔ claude adapter: --restricted is always passed (5.21017ms)
✔ claude adapter: MOSAIC_CLAUDE_BIN wins over PATH and must be an absolute executable (20.159965ms)
✔ claude adapter: a directory pin exits 2 before it runs; a pin with a space runs (5.571463ms)
✔ claude: CLAUDE.md files and auto-memory don't reach the model; without --restricted they do (733.744695ms)
✔ claude: a missing hook or MCP file refuses before claude starts (7.261434ms)
✔ pi: policy tools and typed tools pass, anything else is blocked (4.166813ms)
✔ claude: builtins map from pi names, typed tools need the mcp prefix (0.833573ms)
✔ file tool paths must resolve inside the workspace (1.259141ms)
✔ pi's own path normalisation can't be used to step out (0.93917ms)
✔ a symlink inside the workspace that points out is outside (0.78945ms)
✔ a dangling symlink is refused at any depth, in both harnesses (2.602575ms)
✔ read is checked under every spelling pi's read would open, in both harnesses (14.615549ms)
✔ other spellings cover directories, dangling links and pi's cwd (1.431555ms)
✔ a relative path climbs from the workspace's real path, in both harnesses (4.939839ms)
✔ claude path fields per tool (2.664239ms)
✔ glob patterns stay inside the workspace (1.541148ms)
✔ a path that can't be checked is blocked (0.407067ms)
✔ fetch_review's have lists the working refs before origin's, and no more than HAVE_MAX (205.492658ms)
✔ fetch_review's have orders the working refs newest first within each clone, one clone after another (153.682769ms)
✔ a runner git past its time: SIGTERM reaches its process group, and SIGKILL follows after the grace period (1809.295395ms)
✔ once a signalled runner git has closed, its group gets SIGKILL, so a process that ignored SIGTERM dies too (304.422549ms)
✔ initialize, ping and tools/list (43.470996ms)
✔ tools/call goes through the tool socket; a refusal is an isError result (35.284349ms)
✔ replies are compact JSON; over the cap a tool error names the size and the cap (39.73199ms)
✔ unknown tools and methods are JSON-RPC errors and never reach the socket (30.773489ms)
✔ a missing argument is a usage error (28.366071ms)
✔ pi: typed tools reach the socket, the gate blocks, agent_end writes the marker (390.405442ms)
✔ pi: a reply over the cap is a tool error naming the size and the cap; list_tasks asks for open tasks (337.974176ms)
✔ pi: a write through a dangling symlink is blocked, and nothing appears outside (319.5427ms)
✔ pi: a read is refused when pi would open another spelling outside (317.553243ms)
✔ pi: a relative path climbs from the real path of a workspace behind a symlink (343.7026ms)
✔ pi: a relative path climbs from the real path of a dataRoot behind a symlink (315.863192ms)
✔ pi: a missing extension refuses before any model call (5.557936ms)
✔ pi: an extension without its configuration fails pi's start (251.59832ms)
✔ founderCheck: founder variables, then a needed service without a usable token (1.679202ms)
✔ turnRequest names the sender, class, reply and decision (0.30774ms)
✔ turnRequest says a tester's message is a test and never Jason's instruction (0.331566ms)
✔ turnRequest shows the message's own sent time, a day old as well as new (0.205378ms)
✔ stateBlock names the launch state with its time and event, and the open tasks with the cap (0.321346ms)
✔ a message becomes a turn, the answer goes back as a RESULT, SIGTERM releases and exits 0 (256.202037ms)
✔ launches revoked and then restored: the turn's block reads allowed and gives the restore time (234.616086ms)
✔ a task closed on the tracker is missing from the next turn's open list (274.741324ms)
✔ a message read a day after it was sent shows the time it was sent (171.79872ms)
✔ a failed state read fails the turn: no adapter runs, the sender hears why, the message is read (121.965252ms)
✔ a SIGTERM before the claim stops the runner with exit 0 and no claim (94.691626ms)
✔ typed tools carry the runner's capability; launch goes to the host's launch socket (439.343223ms)
✔ a RESULT gets no automatic reply; failed turns reply with the reason (1303.376946ms)
✔ SIGTERM during a turn kills the turn's process group and still exits 0 (149.352393ms)
✔ founder credentials stop before the claim (20) (170.422241ms)
✔ a refused claim exits 21; an ended run's capability exits 22 (182.195128ms)
✔ the launch ending under a running session exits 22 (143.626848ms)
✔ a broker that stays unreachable exits 23 after brokerRetries polls (239.705116ms)
✔ a broker that is down at the claim exits 23, not 21 (82.223333ms)
✔ no capability, or a malformed one, on stdin exits 2 (182.301051ms)
✔ a missing or malformed policy exits 2 before the claim (124.605908ms)
✔ the PM gets launch, its task verbs and the reads (7.002951ms)
✔ list_requirements reads the requirement ids, for a task.create holder with a tracker (3.548047ms)
✔ a coder gets no launch, no resolve_decision, and no task tools without a tracker (1.501571ms)
✔ a coder with a tracker gets its own task tools and no task_reassign (1.242424ms)
✔ launch only when the business's launch block names the instance as launcher (2.533443ms)
✔ an action outside the instance's authority has no tool (2.09545ms)
✔ the review loop: push_working and request_review for the coder, fetch_review and review_verdict for the reviewer, none for the PM (5.258641ms)
✔ callTool: one JSON line out, the result back, a refusal rejects (7.60395ms)
✔ list_tasks takes the broker's filters and pages; get_task takes a task_ref (2.110825ms)
✔ callTool sends list_tasks done false unless the session names done; other tools as given (2.402655ms)
✔ replyText: compact JSON up to the cap, an error naming the size and the cap over it (0.647169ms)
ℹ tests 80
ℹ suites 0
ℹ pass 80
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 11177.988231