feat(sessions): add exact native continuation

[drewstone]

Outcome

Adds one exact, retry-safe continuation endpoint for retained native sessions.

The endpoint binds caller, operation, request, session, run, and provider boundary before Pi receives another prompt. It stores source-boundary proof before admission and reconstructs a durable terminal result after a settle-window crash.

Proof

• Already up to date
  Done in 271ms using pnpm v11.21.0
  �[33m�[39m
  �[33m ╭───────────────────────────────────────────────╮�[39m
  �[33m│�[39m �[33m│�[39m
  �[33m│�[39m Update available! �[31m11.21.0�[39m → �[32m11.22.0�[39m. �[33m│�[39m
  �[33m│�[39m �[35mChangelog:�[39m https://pnpm.io/v/11.22.0 �[33m│�[39m
  �[33m│�[39m To update, run: �[35mcorepack use pnpm@11.22.0�[39m �[33m│�[39m
  �[33m│�[39m �[33m│�[39m
  �[33m ╰───────────────────────────────────────────────╯�[39m
  �[33m�[39m
• No peer dependency issues found

RUN v2.1.9 /home/drew/code/.worktrees/cli-bridge-native-continuation-20260819

✓ tests/docker-preflight.test.ts (18 tests) 16ms
✓ tests/sse-keepalive.test.ts (15 tests) 9ms
✓ tests/sandbox.test.ts (13 tests) 37ms
✓ tests/scoped-host.test.ts (10 tests | 4 skipped) 24ms
✓ tests/profile-mcp.test.ts (37 tests) 79ms
✓ tests/single-instance.test.ts (11 tests) 633ms
✓ single-instance ownership > rejects a second live bridge process even when it chooses another port 336ms
✓ tests/jail.test.ts (56 tests) 177ms
✓ tests/docker-exec-diagnosis.test.ts (13 tests) 16ms
stdout | tests/codex-backend.test.ts > CodexBackend model translation > accepts an honest unqualified default route
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:804b30d90166d6687d38de5b48ba902a98ed6dfc423225793b163b7315eca73b","harness":"codex","provider":null,"model":"codex/default","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:4d77c3c772a6e4e51147810e5ddf060b6f676dfd8025697db53fd10ce00a7657","files":[],"unsupported":[]}

✓ tests/codex-backend.test.ts (9 tests) 38ms
stdout | tests/materialize-profile.test.ts > materializeProfile — verified per-harness routing > returns and retains a safe receipt for the exact applied workspace plan
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:6c158c618bcb6368b3c71212392b38754d8bb1ebe21884a6c1d721072ceaf86d","harness":"claude-code","provider":null,"model":"claude-code/opus","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:9c1e655e9da34810c098a02c71ed4b71f783a47a37072570dc99e46f72cacfda","files":[{"path":".claude/skills/receipt-skill/SKILL.md","mode":420}],"unsupported":[]}

stdout | tests/materialize-profile.test.ts > materializeProfile — verified per-harness routing > applies Gemini profile systemPrompt exactly once through its native system file
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:e725e571c668f0a29677e1e10918d117a2131d82ca9a86c1f019d4cc0e904afe","harness":"gemini","provider":null,"model":"gemini/gemini-2.5-pro","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:61efc62289abf0f5b12956d127becec0bbe32a77910363813c81a53dfb6f87f9","files":[{"path":".gemini/system.md","mode":420},{"path":"GEMINI.md","mode":420}],"unsupported":[]}

✓ tests/materialize-profile.test.ts (15 tests) 103ms
stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > claude-code > sends a replacement on --system-prompt and never on the additive flag
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:5fcb549c9ef42ca3af1f23086bc6e48b9fac0ac1570b189616b8629824aa2434","harness":"claude-code","provider":null,"model":"claude-code/opus","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:51b9f540e493851f96c55545d567e650abccb8a529ca8627fd3b052aeb48fc80","files":[],"unsupported":[]}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > claude-code > sends an addition on --append-system-prompt and never on the replacement flag
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:a5614ce03930d2508262047da24cddb229eba23398dba0456d121b45cc15360c","harness":"claude-code","provider":null,"model":"claude-code/opus","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:947afbd106167b98fe6fb04bd661714df96f11ec5c77384fc600ebac664b4381","files":[],"unsupported":[]}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > claude-code > keeps the two intents separate when both are set
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:720005eebb0a1899156f0423f456bba799f4d07e8b585024ae3199925e05d8e2","harness":"claude-code","provider":null,"model":"claude-code/opus","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:cb4f0de6ff04d0387e35d399cdd3c79648c3bbb94bd52445b867def7bcce1a32","files":[],"unsupported":[]}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > claude-code > emits --append-system-prompt exactly once when several additive sources apply
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:d12540a06a02cbe564fec9b44a58a280ec97da1b737f401eb2087fe69e1af86c","harness":"claude-code","provider":null,"model":"claude-code/opus","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:713e60a94eaaeae9117cc4eaa88067585b0c401a200e2ecfe271c49e7d0a6580","files":[{"path":".tangle/claude-settings.json","mode":420}],"unsupported":[]}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > pi > writes the replacement to a file on --system-prompt
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:5fcb549c9ef42ca3af1f23086bc6e48b9fac0ac1570b189616b8629824aa2434","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:93ec4bc78a8f31074ffff4a92a9d5206c384546806486bd77e045dbe493c0cd9","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > pi > writes the addition to its own file on --append-system-prompt
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:a5614ce03930d2508262047da24cddb229eba23398dba0456d121b45cc15360c","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:bb1e25e80b8ba5b459953822663bd4397dec86c175b0f9c9fedae7a5d67100d7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > pi > orders the profile addition ahead of the project-instruction file
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:f185b9b721b55d010eca990252ea74c4a59b77992adcf065dba4c7725bd2b988","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:b0db98576fae3987397834aea160d7f6c81eb9cc88cffdbec221e81e8f452628","files":[{"path":"AGENTS.md","mode":420}],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/prompt-intent.test.ts (36 tests) 124ms
stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > codex > points model_instructions_file at the written replacement file
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:5fcb549c9ef42ca3af1f23086bc6e48b9fac0ac1570b189616b8629824aa2434","harness":"codex","provider":null,"model":"codex","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:dd12a5a275010493ff0b2914ab36f09058e5419eec24a645fb1c26c8628b24ad","files":[{"path":".codex/system-prompt.md","mode":420}],"unsupported":[]}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > opencode > registers the addition as its own instructions file, ahead of project instructions
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:f185b9b721b55d010eca990252ea74c4a59b77992adcf065dba4c7725bd2b988","harness":"opencode","provider":null,"model":"opencode","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:6026ffd70b3b1fc75bf50581597a2a0926d012366d3c1c16210f44fe255cad20","files":[{"path":".opencode/agent-system-prompt.md","mode":420},{"path":".opencode/profile-instructions.md","mode":420},{"path":"opencode.json","mode":420}],"unsupported":[]}

stdout | tests/prompt-intent.test.ts > prompt intents reach each harness through its own control > gemini > installs a replacement as its native system file
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:5fcb549c9ef42ca3af1f23086bc6e48b9fac0ac1570b189616b8629824aa2434","harness":"gemini","provider":null,"model":"gemini","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:151e58cff982a13a852c88b1cfc8182810e2fc1ddc2ba1533a4875b86113c6ce","files":[{"path":".gemini/system.md","mode":420}],"unsupported":[]}

✓ tests/protected-model-credential.test.ts (3 tests) 35ms
✓ tests/execution-timeout.test.ts (14 tests) 58ms
✓ tests/admission-lanes.test.ts (8 tests) 420ms
stdout | tests/pi-backend.test.ts > PiBackend turn retry (#125) > exposes only the protected gateway origin in the materialization receipt
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:4e528aea2c4fde95244a09d21bcfd5e595fe025dd206ffa996cb5227149f3fd6","harness":"pi","provider":"tangle-router","model":"pi/tangle-router/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"https://router.tangle.tools","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/acp.test.ts (4 tests) 14ms
✓ tests/claude-cost-receipt.test.ts (5 tests) 27ms
✓ tests/termination-attribution.test.ts (6 tests) 136ms
stdout | tests/pi-backend.test.ts > PiBackend > preserves the provider response model and fingerprint in every emitted delta
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:2d87f61cf56a8d8e41ebb0b34bdb1126e93dce6aa9b19ef1d8d5ad9fd0d86a58","harness":"pi","provider":"tangle-router","model":"pi/tangle-router/deepseek-v4-flash","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-backend.test.ts > PiBackend > applies one exact profile while leaving the task unchanged
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:26ef77a855f4d5fe85ef50b5b7e8718161f5299acb28fce65a7645be1176d5d4","harness":"pi","provider":"tangle-router","model":"pi/tangle-router/glm-5.2","reasoningEffort":{"requested":"ultracode","applied":"xhigh"},"workspacePlanDigest":"sha256:a632e2f06b0e273b7cc8d0c16b102ff35731dc07b09691d3e0277411525d9db6","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/prime-backend.test.ts (47 tests | 1 skipped) 460ms
✓ tests/net-jail.test.ts (50 tests) 924ms
stdout | tests/pi-backend.test.ts > PiBackend > accepts a 'matching max_tokens' and records the profile cap once
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:cf6f41cb466c0dee1b476fc6db62f1467dbfc5b97b90cedbb2240db6c0d0a13c","harness":"pi","provider":"tangle-router","model":"pi/tangle-router/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback","appliedMaxTokens":16384}}

stdout | tests/pi-backend.test.ts > PiBackend > accepts a 'older caller omission' and records the profile cap once
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:cf6f41cb466c0dee1b476fc6db62f1467dbfc5b97b90cedbb2240db6c0d0a13c","harness":"pi","provider":"tangle-router","model":"pi/tangle-router/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback","appliedMaxTokens":16384}}

✓ tests/opencode-worker-timeout.test.ts (7 tests) 19ms
stdout | tests/pi-backend.test.ts > PiBackend > isolates concurrent same-name profile resources and keeps their authority channels distinct
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:24b8e0771eea20ef028b8d5ec7196d377d5a84a398630b707354566f05263eaa","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:91422e5d5fea58b1e083673abbb14282dea4c4eb51a1c4ba8bc775c3e68a02da","files":[{"path":".pi/prompts/review.md","mode":420},{"path":".pi/skills/method/SKILL.md","mode":420},{"path":"AGENTS.md","mode":420}],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:c5e8f58aafd7653bf8f9fbd24c8e4726c9afc95faa56af1369890fb13dc4b9bb","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:bb96350e6e0c41d3c0768def54c71217c5b2bc2bf0c01daa30c2ee2ece6443a6","files":[{"path":".pi/prompts/review.md","mode":420},{"path":".pi/skills/method/SKILL.md","mode":420},{"path":"AGENTS.md","mode":420}],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-backend.test.ts > PiBackend > passes 140k instructions by file path instead of overflowing argv
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:58a8170f767bfabbab822a73b1540943cb3c1c90dff9dd4ad44e34a5a5db5cad","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:780d7b9cc8d8e832ca83cbcb16215eee1dc972f4886a71d5eed79c37ef0bdadd","files":[{"path":"AGENTS.md","mode":420}],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

↓ tests/docker-cancellation-real.test.ts (1 test | 1 skipped)
✓ tests/images-generate.test.ts (6 tests | 1 skipped) 39ms
✓ tests/trace-child-env.test.ts (10 tests) 367ms
↓ tests/pi-native-real.test.ts (4 tests | 4 skipped)
stdout | tests/pi-backend.test.ts > PiBackend > keeps anonymous calls stateless, creates caller sessions, then resumes the mapped Pi session
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:041f13f1162967bf03c879365d73bee759c0aaf93597d7a6a652db5192f96706","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:83b81fe34e834a41f1dd3b1d589c5d6a4ee0b3149936b9039c47652b20743c1c","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback","appliedMaxTokens":16384}}

✓ tests/credential-host-home.test.ts (3 tests) 9ms
✓ tests/model-discovery.test.ts (6 tests) 25ms
✓ tests/durability.test.ts (7 tests) 1127ms
✓ host executor semaphore > blocks the 3rd concurrent spawn until a slot frees 305ms
stdout | tests/pi-backend.test.ts > PiBackend > keeps anonymous calls stateless, creates caller sessions, then resumes the mapped Pi session
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:041f13f1162967bf03c879365d73bee759c0aaf93597d7a6a652db5192f96706","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:83b81fe34e834a41f1dd3b1d589c5d6a4ee0b3149936b9039c47652b20743c1c","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback","appliedMaxTokens":16384}}

✓ tests/execution-router.test.ts (8 tests) 659ms
✓ tests/probe-request-path.test.ts (17 tests) 532ms
✓ tests/run-memory-bounds.test.ts (6 tests) 690ms
✓ durable run memory bounds > does not accumulate unsettled runs across many dispatches 306ms
stdout | tests/resume-rematerialization.test.ts > resume re-materialization > re-applies an agent-edited workspace as a no-op when the session digest matches
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:4fadfed918b26a140a4f51339fe79b1524c58956d64a3370ca6823f95938a9ca","harness":"claude-code","provider":"anthropic","model":"claude-code/anthropic/haiku","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:563c4e49848303d4bd8978720c24ff2ac2d054cdd8bde233b320fe17fe4701d6","files":[{"path":"CLAUDE.md","mode":420}],"unsupported":[]}
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:4fadfed918b26a140a4f51339fe79b1524c58956d64a3370ca6823f95938a9ca","harness":"claude-code","provider":"anthropic","model":"claude-code/anthropic/haiku","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:563c4e49848303d4bd8978720c24ff2ac2d054cdd8bde233b320fe17fe4701d6","files":[{"path":"CLAUDE.md","mode":420}],"unsupported":[]}

stdout | tests/resume-rematerialization.test.ts > resume re-materialization > materializes fully when the session digest was recorded for a different cwd
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:4fadfed918b26a140a4f51339fe79b1524c58956d64a3370ca6823f95938a9ca","harness":"claude-code","provider":"anthropic","model":"claude-code/anthropic/haiku","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:563c4e49848303d4bd8978720c24ff2ac2d054cdd8bde233b320fe17fe4701d6","files":[{"path":"CLAUDE.md","mode":420}],"unsupported":[]}
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:4fadfed918b26a140a4f51339fe79b1524c58956d64a3370ca6823f95938a9ca","harness":"claude-code","provider":"anthropic","model":"claude-code/anthropic/haiku","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:563c4e49848303d4bd8978720c24ff2ac2d054cdd8bde233b320fe17fe4701d6","files":[{"path":"CLAUDE.md","mode":420}],"unsupported":[]}

stdout | tests/resume-rematerialization.test.ts > resume re-materialization > still refuses a mid-session profile swap, naming both content digests
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:4fadfed918b26a140a4f51339fe79b1524c58956d64a3370ca6823f95938a9ca","harness":"claude-code","provider":"anthropic","model":"claude-code/anthropic/haiku","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:563c4e49848303d4bd8978720c24ff2ac2d054cdd8bde233b320fe17fe4701d6","files":[{"path":"CLAUDE.md","mode":420}],"unsupported":[]}

✓ tests/resume-rematerialization.test.ts (3 tests) 50ms
stdout | tests/pi-backend.test.ts > PiBackend > accounts for Google countTokens separately from generation usage receipts
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:5e3f0694d8f4d4a1b238e73fdd89a67634d7ecd61cf871d779bf43b3346f578d","harness":"pi","provider":"google-test","model":"pi/google-test/credential-check","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:41461/v1beta","apiMode":"google-generative-ai","transport":"scoped-loopback"}}

✓ tests/diagnostic-buffer.test.ts (3 tests) 27ms
✓ tests/process-lines.test.ts (3 tests) 105ms
stdout | tests/pi-native.test.ts > Pi native RPC adapter > uses a real JSONL child for two HTTP turns, state proof, and canonical replay
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:e95f60a3471b435e158bf64d0d1389133ff67230b8141571a93ad926f14a2471","harness":"pi","provider":"test","model":"pi/test/model","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:64bf49b9df8d7c95aa717098967353349073bc03d869f6fbe8310ade38aa43aa","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/admission-reviewer-lane.test.ts (4 tests) 812ms
✓ reviewer admission under fleet saturation > BEFORE (reservedActive=0): the fleet fills every slot and the reviewer is rejected 311ms
✓ tests/host-env.test.ts (5 tests) 3ms
✓ tests/opencode-jail-binary.test.ts (3 tests) 13ms
✓ tests/pi-provider-dispatch.test.ts (2 tests) 3ms
✓ tests/nanoclaw.test.ts (5 tests) 169ms
✓ tests/failure-attribution.test.ts (24 tests) 1070ms
✓ tests/docker-failure-honesty.test.ts (33 tests) 1137ms
✓ tests/docker-executor.test.ts (77 tests) 1396ms
stdout | tests/docker-network-wiring.test.ts > Docker server wiring > carries OPENCODE_DOCKER_NETWORK from startup config into the container pool
[opencode-pool] caps memory=4g cpus=2 queue=16 acquire-deadline=60000ms slot-hold=disabled
[opencode-pool] preflight ok on 1 slot(s) — image, credential mounts, HOME, workspace /tmp/cli-bridge-network-wiring-A0A6AU/workspace/opencode, and opencode --version all verified in-slot

stdout | tests/docker-network-wiring.test.ts > Docker server wiring > probes EVERY live slot, each against its own per-slot credential volume
[opencode-pool] caps memory=4g cpus=2 queue=16 acquire-deadline=60000ms slot-hold=disabled
[opencode-pool] preflight ok on 4 slot(s) — image, credential mounts, HOME, workspace /tmp/cli-bridge-perslot-wiring-v9bo2T/workspace/opencode, and opencode --version all verified in-slot

✓ tests/docker-network-wiring.test.ts (2 tests) 148ms
✓ tests/admission-slot-lifetime.test.ts (12 tests) 1198ms
✓ tests/trace-emitter.test.ts (44 tests) 1378ms
✓ tests/server-capabilities.test.ts (1 test) 86ms
✓ tests/mcp-passthrough.test.ts (11 tests) 1597ms
✓ per-backend materializer produces a launchable stdio MCP server > codex config.toml — round-trips through the real codex parser when available 797ms
✓ tests/durable-runs.test.ts (18 tests) 1673ms
stdout | tests/pi-native.test.ts > Pi native RPC adapter > closes the real retained child and removes private files at identity eviction
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:6b57af985dea1687c7f2183aff60b4233780c6c489990e45c9eb738fa98cf17e","harness":"pi","provider":"test","model":"pi/test/model","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e9142f805b08917f4df5cc4598e3bd82ae9492036489bea1ad5eec7a573730b","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/cad-render.test.ts (7 tests) 2505ms
✓ POST /cad/render — real openscad happy path > renders a trivial cube to PNG bytes 1254ms
✓ POST /cad/render — real openscad happy path > soft-fails png while preserving the stl when host has no GL 1193ms
✓ tests/smoke.test.ts (79 tests) 2837ms
stdout | tests/pi-backend.test.ts > PiBackend > an EMPTY extension list means load nothing — the complete-isolation request
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:0a0e702ac3b772150458aec66492c2cb11df9768546f91aa0471a17270ad1b42","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-backend.test.ts > PiBackend > loads an explicit Pi extension set without ambient extension discovery
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:e38f02ceb8bc268638bf1d66455f6ca8c3af6d0805acbb711ca78c800fb0b696","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-backend.test.ts > PiBackend > exposes an exact extension to a read-confined run without copying ambient Pi credentials
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:e38f02ceb8bc268638bf1d66455f6ca8c3af6d0805acbb711ca78c800fb0b696","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-backend.test.ts > PiBackend > removes request-scoped profile files when Pi fails to spawn
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:bab43967bd61b9b06a87eae2240cf66b5c7a90bbdb398085022a4c25775edf90","harness":"pi","provider":"zai-coding-paas","model":"pi/zai-coding-paas/glm-5.2","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:b8b7d2685c88d925190f1ef3d43a2ba154ec8492b219771481d6f0a591cd5131","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:9/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/pi-backend.test.ts (58 tests) 3271ms
✓ PiBackend > rejects invalid Pi token counts instead of recording them as zero 2273ms
✓ tests/pi-native.test.ts (10 tests) 3353ms
✓ Pi native RPC adapter > uses a real JSONL child for two HTTP turns, state proof, and canonical replay 675ms
✓ Pi native RPC adapter > keeps a retained child alive when the Pi RPC timeout is explicitly disabled 422ms
✓ Pi native RPC adapter > closes the real retained child and removes private files at identity eviction 740ms
✓ Pi native RPC adapter > bounds an unacknowledged initial prompt RPC and then closes the real child 312ms
✓ Pi native RPC adapter > acknowledges a Pi interaction only after the native turn visibly resumes 318ms
✓ Pi native RPC adapter > returns transport failure when Pi consumes a response but exposes no resumed native effect 339ms
stdout | tests/pi-inference-isolation.test.ts > Pi inference credential isolation > lowers a profile model cap in real Pi traffic without mutating the operator catalog
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:ab75643757448ff09e59af33677f2dec7163ba18e3455cd4e2e1f16319f5b6bf","harness":"pi","provider":"isolated-test","model":"pi/isolated-test/credential-check","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:2e0962913b2ebd66008754f9574c59b0ff54539a9e92dc161c2c67b5e5d99bf7","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:35923/v1","apiMode":"openai-completions","transport":"scoped-loopback","appliedMaxTokens":64000}}

stdout | tests/pi-inference-isolation.test.ts > Pi inference credential isolation > confines a real Pi Bash turn to its workspace and exact persistent session
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:49a02ddc07933d0b2886f761723cf554614afaf803c166e639442c3918f33f31","harness":"pi","provider":"isolated-test","model":"pi/isolated-test/credential-check","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:62c469faf5b625ff667019b9e6f3a17115f3bbf65e77803c9b523cf9813f670f","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:41741/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-inference-isolation.test.ts > Pi inference credential isolation > confines a real Pi Bash turn to its workspace and exact persistent session
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:49a02ddc07933d0b2886f761723cf554614afaf803c166e639442c3918f33f31","harness":"pi","provider":"isolated-test","model":"pi/isolated-test/credential-check","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:62c469faf5b625ff667019b9e6f3a17115f3bbf65e77803c9b523cf9813f670f","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:41741/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-inference-isolation.test.ts > Pi inference credential isolation > records and refuses an extra model request made directly by a real Pi Bash tool
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:49a02ddc07933d0b2886f761723cf554614afaf803c166e639442c3918f33f31","harness":"pi","provider":"isolated-test","model":"pi/isolated-test/credential-check","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:62c469faf5b625ff667019b9e6f3a17115f3bbf65e77803c9b523cf9813f670f","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:37873/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

stdout | tests/pi-inference-isolation.test.ts > Pi inference credential isolation > rejects a wrong-model request from a real Pi Bash tool and preserves failure evidence
[cli-bridge] profile materialization receipt {"schema":"cli-bridge.profile-materialization.v2","effectiveProfileDigest":"sha256:49a02ddc07933d0b2886f761723cf554614afaf803c166e639442c3918f33f31","harness":"pi","provider":"isolated-test","model":"pi/isolated-test/credential-check","reasoningEffort":{"requested":null,"applied":null},"workspacePlanDigest":"sha256:62c469faf5b625ff667019b9e6f3a17115f3bbf65e77803c9b523cf9813f670f","files":[],"unsupported":[],"inference":{"effectiveEndpoint":"http://127.0.0.1:39583/v1","apiMode":"openai-completions","transport":"scoped-loopback"}}

✓ tests/pi-inference-isolation.test.ts (33 tests | 1 skipped) 8993ms
✓ Pi inference credential isolation > keeps Codex bearer identity outside Pi while accepting its compressed request 414ms
✓ Pi inference credential isolation > atomically reserves one deterministic native file when concurrent first turns race 807ms
✓ Pi inference credential isolation > loads request MCP from the installed adapter in a real read-confined Pi run 2632ms
✓ Pi inference credential isolation > lowers a profile model cap in real Pi traffic without mutating the operator catalog 610ms
✓ Pi inference credential isolation > confines a real Pi Bash turn to its workspace and exact persistent session 1116ms
✓ Pi inference credential isolation > persists an interrupted native first turn before resuming the exact Pi session 1075ms
✓ Pi inference credential isolation > records and refuses an extra model request made directly by a real Pi Bash tool 656ms
✓ Pi inference credential isolation > rejects a wrong-model request from a real Pi Bash tool and preserves failure evidence 668ms
✓ tests/retained-sessions.test.ts (108 tests) 21072ms
✓ retained Agent Interface sessions > binds public provider and environment coordinates into retained turn identity 318ms
✓ retained Agent Interface sessions > keeps retained metadata, context, and provider options separate with deterministic precedence and digesting 319ms
✓ retained Agent Interface sessions > surfaces finalization persistence failure as unknown and refuses continuation 6126ms
✓ retained Agent Interface sessions > survives a real SIGKILL after native effect start without repeating the effect 414ms
✓ retained Agent Interface sessions > accepts native continuation with the exact boundary and supports sequential native continuation turns 312ms
✓ retained Agent Interface sessions > prunes acknowledged interaction records but preserves unknown-effect tombstones 867ms

Test Files 51 passed | 2 skipped (53)
Tests 978 passed | 12 skipped (990)
Start at 08:21:08
Duration 23.97s (transform 13.19s, setup 0ms, collect 58.82s, tests 59.69s, environment 10ms, prepare 5.77s): 978 passed, 12 skipped

• Real provider integration in agent-sdk: one Bridge process, one Pi process, two prompts, one session, exact replay, no third prompt