✔ sessionModel: agent vars win, then the system's execution settings (13.631434ms)
✔ a pi bundle: prompt, policy, tools and manifest, 0600 in a 0700 directory (5.334979ms)
✔ a claude-code bundle adds the wrapped gate hook and the MCP config (2.925694ms)
✔ the prompt's built-in tools line is the harness's own names, as the manifest lists them (11.350753ms)
✔ the prompt names the project clones in the workspace, and none when there are none (7.825398ms)
✔ the prompt explains the review loop to a role with push_working, and reviews to one with fetch_review (6.899916ms)
✔ a bundle is written once: an existing file refuses (1.921315ms)
✔ a path with a single quote can't go into the hook command (2.33337ms)
✔ allow exits 0, a deny exits 2 with the reason on stderr (126.30648ms)
✔ a missing or wrong policy, or a bad event, exits 2 (84.122122ms)
✔ the bundle's wrapped command: a missing gate or node still blocks (1092.925085ms)
✔ claude: typed tools through MCP, the hook blocks, builtins outside --tools don't exist (804.631426ms)
✔ claude: the hook alone blocks a path outside the workspace (471.284235ms)
✔ claude: a second turn resumes the first turn's session (794.447494ms)
✔ claude adapter: --restricted is always passed (5.276069ms)
✔ claude adapter: MOSAIC_CLAUDE_BIN wins over PATH and must be an absolute executable (16.051166ms)
✔ claude adapter: a directory pin exits 2 before it runs; a pin with a space runs (5.495306ms)
✔ claude: CLAUDE.md files and auto-memory don't reach the model; without --restricted they do (702.126561ms)
✔ claude: a missing hook or MCP file refuses before claude starts (6.875762ms)
✔ pi: policy tools and typed tools pass, anything else is blocked (3.151386ms)
✔ claude: builtins map from pi names, typed tools need the mcp prefix (0.81153ms)
✔ file tool paths must resolve inside the workspace (1.494932ms)
✔ pi's own path normalisation can't be used to step out (1.351351ms)
✔ a symlink inside the workspace that points out is outside (1.248688ms)
✔ a dangling symlink is refused at any depth, in both harnesses (4.728526ms)
✔ read is checked under every spelling pi's read would open, in both harnesses (22.086122ms)
✔ other spellings cover directories, dangling links and pi's cwd (1.970961ms)
✔ a relative path climbs from the workspace's real path, in both harnesses (6.817326ms)
✔ claude path fields per tool (0.802524ms)
✔ glob patterns stay inside the workspace (0.749681ms)
✔ a path that can't be checked is blocked (0.482911ms)
✔ fetch_review's have lists the working refs before origin's, and no more than HAVE_MAX (226.273786ms)
✔ fetch_review's have orders the working refs newest first within each clone, one clone after another (176.793169ms)
✔ a runner git past its time: SIGTERM reaches its process group, and SIGKILL follows after the grace period (1810.26278ms)
✔ once a signalled runner git has closed, its group gets SIGKILL, so a process that ignored SIGTERM dies too (304.098196ms)
✔ initialize, ping and tools/list (57.646317ms)
✔ tools/call goes through the tool socket; a refusal is an isError result (32.593553ms)
✔ replies are compact JSON; over the cap a tool error names the size and the cap (36.220775ms)
✔ unknown tools and methods are JSON-RPC errors and never reach the socket (33.349179ms)
✔ a missing argument is a usage error (28.165679ms)
✔ pi: typed tools reach the socket, the gate blocks, agent_end writes the marker (400.22475ms)
✔ pi: a reply over the cap is a tool error naming the size and the cap; list_tasks asks for open tasks (339.431181ms)
✔ pi: a write through a dangling symlink is blocked, and nothing appears outside (333.30204ms)
✔ pi: a read is refused when pi would open another spelling outside (345.905292ms)
✔ pi: a relative path climbs from the real path of a workspace behind a symlink (333.920638ms)
✔ pi: a relative path climbs from the real path of a dataRoot behind a symlink (329.325719ms)
✔ pi: a missing extension refuses before any model call (5.495423ms)
✔ pi: an extension without its configuration fails pi's start (250.090632ms)
✔ founderCheck: founder variables, then a needed service without a usable token (1.725096ms)
✔ turnRequest names the sender, class, reply and decision (0.333692ms)
✔ turnRequest says a tester's message is a test and never Jason's instruction (0.425202ms)
✔ turnRequest shows the message's own sent time, a day old as well as new (0.213431ms)
✔ stateBlock names the launch state with its time and event, and the open tasks with the cap (0.323995ms)
✔ a message becomes a turn, the answer goes back as a RESULT, SIGTERM releases and exits 0 (224.831261ms)
✔ launches revoked and then restored: the turn's block reads allowed and gives the restore time (242.009312ms)
✔ a task closed on the tracker is missing from the next turn's open list (217.669095ms)
✔ a message read a day after it was sent shows the time it was sent (178.086605ms)
✔ a decision the human resolves becomes a turn citing the decision and the answer (238.213198ms)
✔ a failed state read fails the turn: no adapter runs, the sender hears why, the message is read (144.959388ms)
✔ a SIGTERM before the claim stops the runner with exit 0 and no claim (104.934163ms)
✔ typed tools carry the runner's capability; launch goes to the host's launch socket (386.862169ms)
✔ a RESULT gets no automatic reply; failed turns reply with the reason (1301.9623ms)
✔ SIGTERM during a turn kills the turn's process group and still exits 0 (147.612273ms)
✔ founder credentials stop before the claim (20) (169.394611ms)
✔ a refused claim exits 21; an ended run's capability exits 22 (195.147258ms)
✔ the launch ending under a running session exits 22 (145.411067ms)
✔ a broker that stays unreachable exits 23 after brokerRetries polls (254.534966ms)
✔ a broker that is down at the claim exits 23, not 21 (100.963116ms)
✔ no capability, or a malformed one, on stdin exits 2 (201.219652ms)
✔ a missing or malformed policy exits 2 before the claim (131.618187ms)
✔ the PM gets launch, its task verbs and the reads (7.625345ms)
✔ list_requirements reads the requirement ids, for a task.create holder with a tracker (6.328199ms)
✔ a coder gets no launch, no resolve_decision, and no task tools without a tracker (2.049492ms)
✔ a coder with a tracker gets its own task tools and no task_reassign (1.61953ms)
✔ launch only when the business's launch block names the instance as launcher (2.775558ms)
✔ an action outside the instance's authority has no tool (1.338832ms)
✔ the review loop: push_working and request_review for the coder, fetch_review and review_verdict for the reviewer, none for the PM (4.850979ms)
✔ callTool: one JSON line out, the result back, a refusal rejects (11.041601ms)
✔ list_tasks takes the broker's filters and pages; get_task takes a task_ref (3.411336ms)
✔ callTool sends list_tasks done false unless the session names done; other tools as given (3.569702ms)
✔ replyText: compact JSON up to the cap, an error naming the size and the cap over it (1.129709ms)
ℹ tests 81
ℹ suites 0
ℹ pass 81
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 11322.551241