@chankov/agent-fleet 2.1.1 → 2.1.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/.pi/agent-fleet/scripts/lib/fleet-codex-connections.ts +1 -1
- package/.pi/agent-fleet/scripts/lib/fleet-codex-instructions.ts +3 -1
- package/.pi/agent-fleet/scripts/lib/fleet-codex-local-mcp.ts +13 -4
- package/.pi/agent-fleet/scripts/lib/fleet-codex-response-read.ts +26 -0
- package/.pi/agent-fleet/scripts/lib/fleet-codex-response-transport.ts +38 -0
- package/.pi/harnesses/agent-hub/README.md +1 -1
- package/.pi/harnesses/agent-hub/context-continuity-compat.ts +18 -3
- package/.pi/harnesses/agent-hub/package.json +1 -1
- package/.pi/harnesses/coms/package.json +1 -1
- package/.pi/harnesses/damage-control-continue/package.json +1 -1
- package/.pi/harnesses/lib/coms-core.ts +21 -7
- package/.pi/harnesses/lib/coms-response-read.ts +107 -0
- package/.pi/harnesses/lib/fleet-notification-validation.ts +4 -3
- package/.pi/harnesses/lib/fleet-transcript-store.ts +1 -1
- package/.versions/2.1.2/.pi/agent-fleet/hermes/README.md +347 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/desktop-plugins/agent-fleet-herdr/plugin.js +1291 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/desktop-plugins/agent-fleet-herdr/presentation.js +672 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/desktop-plugins/agent-fleet-monitor/plugin.js +19 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/desktop-plugins/agent-fleet-monitor/state.js +10 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/gates/gate-o-capability.md +8 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/gates/gate-o-validator-fixtures.json +1 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/activity.py +773 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/coms_registry.py +134 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/coms_send.py +350 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/herdr_source.py +193 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/manifest.json +4 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/plugin_api.py +444 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/sessions.py +205 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/tasks.py +300 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/dashboard/watch.py +819 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-herdr/plugin.yaml +8 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-monitor/dashboard/adapter.py +139 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-monitor/dashboard/manifest.json +11 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/plugins/agent-fleet-monitor/dashboard/plugin_api.py +50 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-conductor/SKILL.md +84 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-liaison/SKILL.md +84 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/SKILL.md +76 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/manifest.json +1 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog.py +249 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_actions.py +124 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_contract.py +76 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_delivery.py +10 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_engine.py +128 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_judgment.py +41 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_recovery.py +103 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_state.py +163 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/skills/hub-watchdog/scripts/watchdog_transport.py +138 -0
- package/.versions/2.1.2/.pi/agent-fleet/hermes/watchdog-tests/__init__.py +0 -0
- package/.versions/2.1.2/.pi/agent-fleet/hooks/coms-stop-hook.mjs +84 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/claude-code-preflight.ts +70 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/coms-claude-bridge.ts +440 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/coms-cli.ts +430 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/coms-hermes-bridge.ts +416 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/exec-at-target.ts +10 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/fixtures/fleet-codex-allowlist-store-child.ts +31 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/fixtures/response-fixture.ts +23 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/fleet-codex-local-mcp.mjs +26 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/fleet-codex-local-mcp.ts +59 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/fleet.ts +248 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/flow.ts +166 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/hermes-monitor-capability-runner.ts +25 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/hermes-monitor-scenario.ts +474 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/hermes-watchdog-capability-probe.ts +45 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/install-hermes-plugin.sh +180 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/claude-bridge-core.ts +208 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/codex-desktop-queue-host.ts +106 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/codex-unix-queue-rpc.ts +91 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/coms-envelope.ts +582 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fake-herdr-server.ts +88 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-allowlist-store.ts +367 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-allowlist.ts +175 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-answer-transport.ts +73 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-answers.ts +74 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-connections.ts +70 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-event-reply.ts +58 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-event-runtime.ts +116 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-instruction-store.ts +428 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-instruction-transport.ts +160 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-instructions.ts +175 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-local-mcp.ts +415 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-native-history.ts +215 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-observation-store.ts +208 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-observe-transport.ts +49 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-observe.ts +93 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-overview.ts +52 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-permission-transport.ts +23 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-permissions.ts +83 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-readonly.ts +213 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-relay-latency.ts +45 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-response-read.ts +26 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-response-transport.ts +38 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-codex-resync.ts +212 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/fleet-command.ts +268 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/flow-command.ts +107 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/flow-maintenance.ts +238 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/herdr-layout.ts +281 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-bridge-core.ts +241 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-monitor-events.ts +1 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-monitor-herdr.ts +2 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-monitor-model.ts +3 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-monitor-registry.ts +2 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-monitor-socket.ts +2 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/hermes-monitor-store.ts +2 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/monitor-env.ts +146 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/peer-launch.ts +318 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/peer-target.ts +1 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/persona-banner.ts +62 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/persona-dirs.ts +25 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/resource-closure.js +1 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/runtime-dependencies.js +124 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/session-launch.ts +56 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/spawn-stagger.ts +101 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/team-project.ts +178 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/lib/team-snapshot.ts +160 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/package-lock.json +57 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/package.json +10 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/peer-banner.ts +41 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/peer-launch.ts +289 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/pi-session.ts +9 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/team-snapshot.ts +245 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/team-up.ts +385 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/agent-phase.ts +181 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/changes.ts +61 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/debate.ts +149 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/env.ts +36 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/envelopes.ts +169 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/gates.ts +81 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/git.ts +99 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/isolated-snapshot.ts +130 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/merge.ts +95 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/model-profile.ts +22 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/model-visibility.ts +92 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/permissions.ts +137 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/personas.ts +57 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/poll.ts +86 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/project-policy.ts +37 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/quality.ts +65 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/run.ts +109 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/scout-data-boundary.ts +55 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/trace.ts +52 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/voices.ts +128 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/writer-lease-race.worker.mjs +36 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/lib/writer-lease.ts +148 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/tsconfig.json +22 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/wf-build-test.ts +61 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/wf-debate.ts +87 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/wf-document.ts +50 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/wf-poll.ts +95 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/wf-quality.ts +20 -0
- package/.versions/2.1.2/.pi/agent-fleet/scripts/workflows/wf-scout.ts +40 -0
- package/.versions/2.1.2/.pi/agents/dispatch-policy.yaml +32 -0
- package/.versions/2.1.2/.pi/agents/model-profiles.yaml +134 -0
- package/.versions/2.1.2/.pi/agents/peers.yaml +147 -0
- package/.versions/2.1.2/.pi/agents/teams.yaml +41 -0
- package/.versions/2.1.2/.pi/agents/voices.yaml +13 -0
- package/.versions/2.1.2/.pi/damage-control-rules.yaml +297 -0
- package/.versions/2.1.2/.pi/extensions/btw/README.md +120 -0
- package/.versions/2.1.2/.pi/extensions/btw/index.ts +916 -0
- package/.versions/2.1.2/.pi/extensions/btw/package.json +6 -0
- package/.versions/2.1.2/.pi/extensions/btw/state.js +179 -0
- package/.versions/2.1.2/.pi/extensions/chrome-devtools-mcp/README.md +55 -0
- package/.versions/2.1.2/.pi/extensions/chrome-devtools-mcp/index.ts +93 -0
- package/.versions/2.1.2/.pi/extensions/chrome-devtools-mcp/package.json +6 -0
- package/.versions/2.1.2/.pi/extensions/compact-and-continue/README.md +42 -0
- package/.versions/2.1.2/.pi/extensions/compact-and-continue/index.ts +136 -0
- package/.versions/2.1.2/.pi/extensions/compact-and-continue/package.json +6 -0
- package/.versions/2.1.2/.pi/extensions/mcp-bridge/README.md +46 -0
- package/.versions/2.1.2/.pi/extensions/mcp-bridge/index.ts +206 -0
- package/.versions/2.1.2/.pi/extensions/mcp-bridge/package.json +6 -0
- package/.versions/2.1.2/.pi/extensions/package-lock.json +1175 -0
- package/.versions/2.1.2/.pi/extensions/package.json +9 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/README.md +224 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/config.ts +199 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/examples/stt.json +35 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/index.ts +248 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/package.json +6 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/providers.ts +214 -0
- package/.versions/2.1.2/.pi/extensions/pi-voice-stt/recorder.ts +158 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/README.md +1584 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/acceptance.ts +204 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/access-approval.ts +266 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/action-authorization.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-contract.ts +43 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-evidence.ts +64 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-hub.ts +55 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-runtime.ts +56 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-sources-worker.mjs +4 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-sources.ts +74 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/agentic-state.ts +23 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/approval-wait.ts +1 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/artifacts.js +146 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/assertion-ledger.js +68 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/assist-profile.ts +19 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/backend-policy.js +174 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/bounded-output.ts +26 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/budget-recovery.ts +123 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/capability-confirmation.ts +47 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/capability-packs.ts +184 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agent-model-thinking.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agent-model.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agent-models-substitute.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-add.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-drop.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-history.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-kill.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-list.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-restart.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-save.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/agents-team.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/audit.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/compound.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/coms.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/config.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/context-command.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/context.ts +47 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/debate.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/dispatch-policy.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/handoff.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/hub-report.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/models.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/poll.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/recover.ts +61 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/retry.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/task-triage-recover.ts +12 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/task-triage-waive.ts +12 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/watchdog.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/work-mode.ts +29 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/commands/zoom.ts +10 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/config/agents.ts +181 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/config/model-profiles.ts +237 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/config/overrides.ts +166 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context/agent-state.ts +30 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context/assertions-artifacts.ts +221 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context/budgets.ts +152 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context/hub-state.ts +40 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-budget-snapshot.ts +105 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-compaction-monitor.ts +35 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-compaction-preflight.ts +21 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-compaction-runtime.ts +69 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-compaction.ts +34 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity-broker.ts +44 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity-child.ts +83 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity-client.ts +22 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity-compat.ts +45 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity-owners.ts +17 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity-registry.ts +8 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-continuity.ts +82 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-pressure.ts +222 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/context-window.js +108 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-approval.ts +83 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-broker.ts +125 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-contract.ts +22 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-descendant.ts +76 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-hub.ts +263 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-identity.ts +12 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-owners.ts +59 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-state.ts +43 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/damage-guard-test-tsconfig.json +19 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/debate-command.ts +176 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/delegate.ts +478 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/deterministic-fs.ts +195 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/diagnostic-probe-extension.ts +26 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/diagnostic-series-budget.ts +178 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-coms.ts +258 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-core.ts +27 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-native-complete.ts +176 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-native-prepare.ts +301 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-native-spawn.ts +262 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-native-types.ts +293 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-native.ts +236 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-observability.ts +225 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-triage-contract.ts +27 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-triage-eval.ts +77 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-triage-policy.ts +29 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-triage-runtime.ts +55 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/dispatch-triage-state.ts +33 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/docs-lane.js +123 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/drift-judge.ts +352 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/drift-runtime.ts +480 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/drift-system1-policy.ts +39 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/drift-system1.ts +469 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/drift-watchdog.js +349 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/evidence-manifest.ts +51 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/evidence-rules.js +96 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/execution-evidence.ts +96 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/execution-trace.ts +241 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/external-blocker.js +92 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/adapters.ts +88 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/broker-client.ts +62 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/broker.ts +128 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/child-extension.ts +42 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/descendant.ts +67 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/discovery-worker.mjs +15 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/discovery.ts +126 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/evaluate.ts +57 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/hub.ts +170 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/native-context.ts +8 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/native-fixture.test-support.ts +108 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/owners.ts +100 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/results.ts +55 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/runtime.ts +130 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/semantic-profiles.ts +18 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/sources.ts +69 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/file-discovery/state.ts +22 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/filesystem-tool.ts +166 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/fixtures/registration-surface.json +68 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/fixtures/t0a-synthetic/catalog.ts +189 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/footer.ts +30 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/helpers.ts +270 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/history-projection.ts +116 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/hub-state-audit.js +41 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/index.ts +2705 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/input/completions.ts +82 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/input/shortcuts.ts +23 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/lifecycle/context-pressure.ts +231 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/lifecycle/monitor-session.ts +104 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/lifecycle/session-orchestration.ts +145 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/lifecycle/turn-handlers.ts +90 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-control.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-invoke-journal.ts +131 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-invoke.ts +92 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-lifecycle.ts +13 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-publisher.ts +59 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-recovery.ts +3 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-runtime.ts +3 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/monitor-session-bridge.ts +266 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/no-progress.ts +622 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/operator-current-state.ts +53 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/operator-watchdog.ts +97 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/orchestration-activation.ts +109 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/orchestration-metrics.ts +78 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/orchestration-route.ts +68 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/package.json +7 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/peer-spawn-plan.ts +116 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/permission-channel.ts +87 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/permission-relay.ts +84 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/policy/models.ts +153 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/policy/profile-activation.ts +62 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/policy/profile-runtime.ts +120 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/policy/roster.ts +94 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/policy/work-mode.ts +142 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/poll-command.ts +176 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/presentation.ts +54 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/private-config-current-state.ts +81 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/private-config-fixtures.ts +26 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/private-config-review.ts +100 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-config.ts +22 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-evaluate.ts +138 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-external-git.ts +82 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-external.ts +142 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-feedback.ts +104 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-findings.ts +115 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-local.ts +89 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-observer.ts +106 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-rules.ts +171 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-runtime.ts +318 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-selection.ts +68 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-snapshot-worker.mjs +12 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-snapshot.ts +279 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/proactive-types.ts +70 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/process-obligations.ts +177 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/prompts/context.ts +55 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/prompts/fragments.ts +86 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/prompts/session-start.ts +90 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/prompts/system-prompt.ts +125 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/provider-semaphore.js +138 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/reconcile.ts +26 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/recover-policy.ts +44 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/recover-state.ts +104 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/recovery-contract.ts +151 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/rerun-policy.ts +40 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/research/controls.ts +74 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/research/runtime.ts +215 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/research/spawn-run.ts +221 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/research-watchdog.ts +40 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/return-contract.js +371 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/return-extract.js +77 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/review-closure.ts +43 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/review-continuity.ts +73 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/review-findings.js +96 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/review-reuse.ts +47 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/review-verdict.ts +6 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/routing-observation.ts +41 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/routing-proposal.ts +46 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/run-budget.js +342 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/run-namespace.js +93 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/runtime-receipts.ts +63 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/runtime-test-check.ts +85 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/safety-routing.ts +28 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/scope-gate.js +233 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/session-audit.ts +331 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/session-health.js +110 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/session-start.ts +42 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/spawn.ts +702 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/specialist-result-receipt.ts +13 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/specialist-result.ts +78 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/system1-activity.ts +625 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/system1-communication-store.ts +225 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/system1-report.ts +275 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/system1-runtime.ts +271 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-continuation.ts +4 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-resume-contract.ts +80 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-authorization.ts +170 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-config.ts +21 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-contract.ts +14 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-live-stage-boundary.mjs +49 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-obligations.ts +120 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-policy.ts +16 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-review-contract.mjs +43 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-reviewer-only-contract.mjs +156 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-reviewer-only-diagnostics.mjs +112 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-runtime.ts +130 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-stage-guard.ts +21 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/task-triage-state.ts +46 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/timeline.ts +72 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tool-catalog-runtime.ts +79 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tool-catalog-state.ts +123 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tool-protocol.ts +134 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/action-executors.ts +216 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/ask-system1-files.ts +15 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/ask-system1.ts +25 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/coms-tools.ts +139 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/context.ts +147 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/dispatch-agent.ts +92 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/dispatch-execution.ts +777 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/dispatch-triage.ts +16 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/execution-orchestration.ts +22 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/fleet-tools.ts +201 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/herdr-executors.ts +100 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/run-flow.ts +146 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/set-task-tier.ts +31 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/spawn-research.ts +67 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/team-adjust.ts +29 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tools/verification-contract.ts +81 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/tsconfig.json +22 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/types.ts +66 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/activity-dots.ts +25 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/context-budget.ts +89 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/detail-panel.ts +192 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/fleet-actions.ts +78 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/fleet-dashboard.ts +163 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/fleet-editor.ts +44 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/fleet-source.ts +179 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/grid.ts +159 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/history-store.ts +137 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/history.ts +234 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/pool.ts +103 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/running-strip.ts +2 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/system1-communication.ts +59 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/ui/zoom.ts +206 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/unknown-tool-counter.ts +116 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/version.ts +42 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/watchdog-evaluation.ts +110 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/work-mode-controls.ts +56 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/work-mode.ts +188 -0
- package/.versions/2.1.2/.pi/harnesses/agent-hub/write-isolation.ts +187 -0
- package/.versions/2.1.2/.pi/harnesses/ask-user-remote/README.md +59 -0
- package/.versions/2.1.2/.pi/harnesses/ask-user-remote/index.ts +571 -0
- package/.versions/2.1.2/.pi/harnesses/ask-user-remote/questions.ts +206 -0
- package/.versions/2.1.2/.pi/harnesses/ask-user-remote/race-core.js +124 -0
- package/.versions/2.1.2/.pi/harnesses/ask-user-remote/runtime-ask.ts +90 -0
- package/.versions/2.1.2/.pi/harnesses/coms/README.md +105 -0
- package/.versions/2.1.2/.pi/harnesses/coms/index.ts +308 -0
- package/.versions/2.1.2/.pi/harnesses/coms/package.json +7 -0
- package/.versions/2.1.2/.pi/harnesses/coms/version.ts +42 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/README.md +161 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/index.ts +659 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/package.json +7 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/result-screen.ts +98 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/system1-child.ts +45 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/system1-client.ts +85 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/system1-guard.ts +65 -0
- package/.versions/2.1.2/.pi/harnesses/damage-control-continue/version.ts +42 -0
- package/.versions/2.1.2/.pi/harnesses/lib/approval-wait.ts +17 -0
- package/.versions/2.1.2/.pi/harnesses/lib/bash-policy-analysis.ts +185 -0
- package/.versions/2.1.2/.pi/harnesses/lib/changed-file-diagnostics.ts +642 -0
- package/.versions/2.1.2/.pi/harnesses/lib/clarification-policy.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/lib/coms-core-io.ts +422 -0
- package/.versions/2.1.2/.pi/harnesses/lib/coms-core.ts +734 -0
- package/.versions/2.1.2/.pi/harnesses/lib/coms-instruction-receipts.ts +113 -0
- package/.versions/2.1.2/.pi/harnesses/lib/coms-observation.ts +74 -0
- package/.versions/2.1.2/.pi/harnesses/lib/coms-registry-entry.ts +99 -0
- package/.versions/2.1.2/.pi/harnesses/lib/coms-response-read.ts +107 -0
- package/.versions/2.1.2/.pi/harnesses/lib/context-budget-child-prompt.ts +214 -0
- package/.versions/2.1.2/.pi/harnesses/lib/context-budget-hub-prompt.ts +121 -0
- package/.versions/2.1.2/.pi/harnesses/lib/context-budget-view.ts +112 -0
- package/.versions/2.1.2/.pi/harnesses/lib/context-budget.ts +114 -0
- package/.versions/2.1.2/.pi/harnesses/lib/context-compaction-policy.js +35 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-action-hash.ts +28 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-auto-policy.ts +52 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-control-presentation.ts +14 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-control-shared.ts +276 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-guard-port.ts +25 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-policy.ts +35 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-result-evidence.ts +25 -0
- package/.versions/2.1.2/.pi/harnesses/lib/damage-result-view.ts +16 -0
- package/.versions/2.1.2/.pi/harnesses/lib/deterministic-handle-path.ts +25 -0
- package/.versions/2.1.2/.pi/harnesses/lib/exact-permission-port.ts +38 -0
- package/.versions/2.1.2/.pi/harnesses/lib/execution-trace.ts +45 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fixtures/context-budget-profiles.json +11 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fixtures/hub-prompt-operator.legacy.txt +73 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fixtures/hub-prompt-orchestrator.legacy.txt +73 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-codex-event-bridge.ts +82 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-config/cli-bridge.ts +111 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-config/mutations.ts +201 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-config/snapshot.ts +417 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-config/transition.ts +68 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-config/ui.ts +269 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-config/view.ts +338 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-connection-control.ts +77 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-dashboard-ops.ts +178 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-dashboard-view.ts +174 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-detail-view.ts +339 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-event-source.ts +21 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-notification-validation.ts +67 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-overlay.ts +26 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-panel.ts +34 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-question-reference.ts +42 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-read-model.ts +609 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-selection.ts +24 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-strip-controller.ts +75 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-strip-view.ts +179 -0
- package/.versions/2.1.2/.pi/harnesses/lib/fleet-transcript-store.ts +244 -0
- package/.versions/2.1.2/.pi/harnesses/lib/herdr-client.ts +472 -0
- package/.versions/2.1.2/.pi/harnesses/lib/herdr-presence.ts +399 -0
- package/.versions/2.1.2/.pi/harnesses/lib/hermes-monitor-events.ts +136 -0
- package/.versions/2.1.2/.pi/harnesses/lib/hermes-monitor-herdr.ts +28 -0
- package/.versions/2.1.2/.pi/harnesses/lib/hermes-monitor-model.ts +18 -0
- package/.versions/2.1.2/.pi/harnesses/lib/hermes-monitor-registry.ts +6 -0
- package/.versions/2.1.2/.pi/harnesses/lib/hermes-monitor-socket.ts +163 -0
- package/.versions/2.1.2/.pi/harnesses/lib/hermes-monitor-store.ts +31 -0
- package/.versions/2.1.2/.pi/harnesses/lib/identical-dialog.ts +27 -0
- package/.versions/2.1.2/.pi/harnesses/lib/installation-state.js +62 -0
- package/.versions/2.1.2/.pi/harnesses/lib/peer-target.ts +24 -0
- package/.versions/2.1.2/.pi/harnesses/lib/policy-roots.js +117 -0
- package/.versions/2.1.2/.pi/harnesses/lib/policy-roots.ts +32 -0
- package/.versions/2.1.2/.pi/harnesses/lib/process-shared.ts +9 -0
- package/.versions/2.1.2/.pi/harnesses/lib/resource-closure.js +26 -0
- package/.versions/2.1.2/.pi/harnesses/lib/safe-source-read.js +29 -0
- package/.versions/2.1.2/.pi/harnesses/lib/semantic-approval.ts +80 -0
- package/.versions/2.1.2/.pi/harnesses/lib/session-target.ts +62 -0
- package/.versions/2.1.2/.pi/harnesses/lib/spawned-peers.js +191 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-agentic.js +15 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-context-compaction.js +16 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-damage-guard.js +24 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-file-discovery.js +23 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-json.js +37 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-loader.js +53 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-proactive.js +68 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-triage.js +42 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config-v2.js +112 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/config.js +50 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/contracts.ts +154 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/demo.ts +95 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/jev.ts +503 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/payload-screen.ts +65 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/selection.js +11 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/service.ts +245 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/worktree-scopes.js +110 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/worktree-scopes.ts +13 -0
- package/.versions/2.1.2/.pi/harnesses/lib/system1/worktree-service.ts +24 -0
- package/.versions/2.1.2/.pi/harnesses/lib/target-capabilities.js +7 -0
- package/.versions/2.1.2/.pi/harnesses/lib/update-check.ts +227 -0
- package/.versions/2.1.2/.pi/harnesses/lib/workspace-context.js +115 -0
- package/.versions/2.1.2/.pi/harnesses/lib/workspace-context.ts +22 -0
- package/.versions/2.1.2/.pi/harnesses/package-lock.json +2121 -0
- package/.versions/2.1.2/.pi/harnesses/package.json +13 -0
- package/.versions/2.1.2/.pi/prompts/af-build.md +27 -0
- package/.versions/2.1.2/.pi/prompts/af-code-simplify.md +22 -0
- package/.versions/2.1.2/.pi/prompts/af-constraints.md +31 -0
- package/.versions/2.1.2/.pi/prompts/af-plan.md +17 -0
- package/.versions/2.1.2/.pi/prompts/af-review.md +16 -0
- package/.versions/2.1.2/.pi/prompts/af-set-hermes-telegram.md +15 -0
- package/.versions/2.1.2/.pi/prompts/af-set-hermes-watchdog.md +6 -0
- package/.versions/2.1.2/.pi/prompts/af-setup-rules.md +20 -0
- package/.versions/2.1.2/.pi/prompts/af-ship.md +17 -0
- package/.versions/2.1.2/.pi/prompts/af-spec.md +15 -0
- package/.versions/2.1.2/.pi/prompts/af-test.md +19 -0
- package/.versions/2.1.2/.pi/skills/bowser/SKILL.md +145 -0
- package/.versions/2.1.2/.version +1 -0
- package/.versions/2.1.2/agents/architect.md +24 -0
- package/.versions/2.1.2/agents/bowser.md +28 -0
- package/.versions/2.1.2/agents/builder.md +55 -0
- package/.versions/2.1.2/agents/code-reviewer.md +209 -0
- package/.versions/2.1.2/agents/deep-researcher.md +32 -0
- package/.versions/2.1.2/agents/documenter.md +22 -0
- package/.versions/2.1.2/agents/orchestrator.md +22 -0
- package/.versions/2.1.2/agents/plan-reviewer.md +83 -0
- package/.versions/2.1.2/agents/planner.md +107 -0
- package/.versions/2.1.2/agents/releaser.md +24 -0
- package/.versions/2.1.2/agents/researcher.md +30 -0
- package/.versions/2.1.2/agents/security-auditor.md +167 -0
- package/.versions/2.1.2/agents/test-engineer.md +153 -0
- package/.versions/2.1.2/agents/web-debugger.md +44 -0
- package/.versions/2.1.2/agents/web-performance-auditor.md +194 -0
- package/.versions/2.1.2/bin/catalog/harness-runtime-closure.json +85 -0
- package/.versions/2.1.2/docs/ARCHITECTURE.md +1195 -0
- package/.versions/2.1.2/docs/agent-fleet-setup.md +1143 -0
- package/.versions/2.1.2/docs/claude-code-coms-bridge.md +147 -0
- package/.versions/2.1.2/docs/codex-local-mcp.md +503 -0
- package/.versions/2.1.2/docs/codex-session-bridge.md +44 -0
- package/.versions/2.1.2/docs/codex-setup.md +183 -0
- package/.versions/2.1.2/docs/coms-hermes-bridge.md +188 -0
- package/.versions/2.1.2/docs/fleet-native-voice.md +145 -0
- package/.versions/2.1.2/docs/getting-started.md +126 -0
- package/.versions/2.1.2/docs/npm-install.md +535 -0
- package/.versions/2.1.2/docs/pi-extensions.md +455 -0
- package/.versions/2.1.2/docs/pi-setup.md +92 -0
- package/.versions/2.1.2/docs/skill-anatomy.md +150 -0
- package/.versions/2.1.2/install-manifest.json +3239 -0
- package/.versions/2.1.2/justfile +247 -0
- package/.versions/2.1.2/manifest-meta.json +1087 -0
- package/.versions/2.1.2/references/accessibility-checklist.md +160 -0
- package/.versions/2.1.2/references/communication-contract.md +26 -0
- package/.versions/2.1.2/references/definition-of-done.md +67 -0
- package/.versions/2.1.2/references/fleet-coordination-patterns.md +171 -0
- package/.versions/2.1.2/references/observability-checklist.md +92 -0
- package/.versions/2.1.2/references/orchestration-patterns.md +370 -0
- package/.versions/2.1.2/references/performance-checklist.md +236 -0
- package/.versions/2.1.2/references/prompting-patterns.md +380 -0
- package/.versions/2.1.2/references/security-checklist.md +245 -0
- package/.versions/2.1.2/references/testing-patterns.md +235 -0
- package/.versions/2.1.2/references/workflow-authoring.md +304 -0
- package/.versions/2.1.2/skills/_internal/grilling.md +59 -0
- package/.versions/2.1.2/skills/browser-testing-with-devtools/SKILL.md +351 -0
- package/.versions/2.1.2/skills/code-review-and-quality/SKILL.md +396 -0
- package/.versions/2.1.2/skills/compound-learning/SKILL.md +137 -0
- package/.versions/2.1.2/skills/constraint-driven-development/SKILL.md +338 -0
- package/.versions/2.1.2/skills/constraint-driven-development/references/floor-guard.md +99 -0
- package/.versions/2.1.2/skills/context-engineering/SKILL.md +341 -0
- package/.versions/2.1.2/skills/deprecation-and-migration/SKILL.md +247 -0
- package/.versions/2.1.2/skills/designing-agents/SKILL.md +394 -0
- package/.versions/2.1.2/skills/designing-agents/pi-harness-authoring.md +222 -0
- package/.versions/2.1.2/skills/drafting-workflows/SKILL.md +143 -0
- package/.versions/2.1.2/skills/drafting-workflows/scripts/make-workflow.ts +177 -0
- package/.versions/2.1.2/skills/fleet-connect/SKILL.md +25 -0
- package/.versions/2.1.2/skills/fleet-native-voice/SKILL.md +87 -0
- package/.versions/2.1.2/skills/fleet-native-voice/agents/openai.yaml +4 -0
- package/.versions/2.1.2/skills/fleet-native-voice.zip +0 -0
- package/.versions/2.1.2/skills/frontend-ui-engineering/SKILL.md +328 -0
- package/.versions/2.1.2/skills/git-workflow-and-versioning/SKILL.md +316 -0
- package/.versions/2.1.2/skills/idea-refine/SKILL.md +182 -0
- package/.versions/2.1.2/skills/idea-refine/examples.md +238 -0
- package/.versions/2.1.2/skills/idea-refine/frameworks.md +99 -0
- package/.versions/2.1.2/skills/idea-refine/refinement-criteria.md +113 -0
- package/.versions/2.1.2/skills/idea-refine/scripts/idea-refine.sh +15 -0
- package/.versions/2.1.2/skills/incremental-implementation/SKILL.md +353 -0
- package/.versions/2.1.2/skills/incremental-implementation.zip +0 -0
- package/.versions/2.1.2/skills/interview-me/SKILL.md +226 -0
- package/.versions/2.1.2/skills/orchestration-verification/SKILL.md +182 -0
- package/.versions/2.1.2/skills/peer-coms/SKILL.md +90 -0
- package/.versions/2.1.2/skills/performance-optimization/SKILL.md +496 -0
- package/.versions/2.1.2/skills/planning-and-task-breakdown/SKILL.md +318 -0
- package/.versions/2.1.2/skills/planning-and-task-breakdown.zip +0 -0
- package/.versions/2.1.2/skills/repository-ai-setup/SKILL.md +135 -0
- package/.versions/2.1.2/skills/repository-ai-setup/references/agent-fleet-ai-state.template.json +4 -0
- package/.versions/2.1.2/skills/repository-ai-setup/references/apply-contract.md +74 -0
- package/.versions/2.1.2/skills/repository-ai-setup/references/provenance-template.md +16 -0
- package/.versions/2.1.2/skills/repository-ai-setup/references/setup-flow.md +108 -0
- package/.versions/2.1.2/skills/repository-ai-setup/references/verify-setup.md +20 -0
- package/.versions/2.1.2/skills/repository-ai-setup.zip +0 -0
- package/.versions/2.1.2/skills/security-and-hardening/SKILL.md +524 -0
- package/.versions/2.1.2/skills/spec-driven-development/SKILL.md +295 -0
- package/.versions/2.1.2/skills/spec-driven-development.zip +0 -0
- package/.versions/2.1.2/skills/using-agent-skills/SKILL.md +198 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/LICENSE +21 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/README.md +430 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/api-and-interface-design/SKILL.md +367 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/browser-testing-with-devtools/SKILL.md +317 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/ci-cd-and-automation/SKILL.md +390 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/code-review-and-quality/SKILL.md +396 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/code-simplification/SKILL.md +331 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/constraint-driven-development/SKILL.md +311 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/constraint-driven-development/references/floor-guard.md +99 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/context-engineering/SKILL.md +337 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/debugging-and-error-recovery/SKILL.md +300 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/deprecation-and-migration/SKILL.md +247 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/documentation-and-adrs/SKILL.md +288 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/doubt-driven-development/SKILL.md +243 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/frontend-ui-engineering/SKILL.md +328 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/git-workflow-and-versioning/SKILL.md +355 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/idea-refine/SKILL.md +178 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/idea-refine/examples.md +238 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/idea-refine/frameworks.md +99 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/idea-refine/refinement-criteria.md +113 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/idea-refine/scripts/idea-refine.sh +15 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/incremental-implementation/SKILL.md +249 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/interview-me/SKILL.md +225 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/observability-and-instrumentation/SKILL.md +238 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/performance-optimization/SKILL.md +496 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/planning-and-task-breakdown/SKILL.md +257 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/security-and-hardening/SKILL.md +524 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/shipping-and-launch/SKILL.md +330 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/source-driven-development/SKILL.md +216 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/spec-driven-development/SKILL.md +245 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/test-driven-development/SKILL.md +398 -0
- package/.versions/2.1.2/vendor/agent-skills-upstream/skills/using-agent-skills/SKILL.md +192 -0
- package/CHANGELOG.md +7 -0
- package/docs/ARCHITECTURE.md +8 -0
- package/docs/codex-local-mcp.md +16 -0
- package/docs/fleet-native-voice.md +8 -0
- package/docs/system1-context-continuity.md +11 -5
- package/install-manifest.json +1 -1
- package/package.json +1 -1
- package/skills/fleet-native-voice.zip +0 -0
|
@@ -0,0 +1,503 @@
|
|
|
1
|
+
# Native Codex MCP for Agent Fleet
|
|
2
|
+
|
|
3
|
+
Use the local `af-local-read` MCP server to work with **already running** Fleet/Pi sessions from one Codex conversation. Despite its historical name, the server supports permission-controlled observations, instructions, and human question answers across multiple explicitly selected targets.
|
|
4
|
+
|
|
5
|
+
The optional [native chat connection workflow](codex-connections.md) adds exact-target preview/commit, default 12-hour leases for new connections, in-place renewal of an already configured notification mapping, and a Fleet View chat-status panel. First-time notification bootstrap and automatic restart recovery remain explicit blockers.
|
|
6
|
+
|
|
7
|
+
This guide covers the current native interface. The [legacy CLI client](codex-session-bridge.md) is a separate interface with separate state; its selection and receipts are not imported automatically. The optional [native voice workflow](fleet-native-voice.md) uses the same MCP tools and permissions, not a different transport.
|
|
8
|
+
|
|
9
|
+
## Requirements and support boundaries
|
|
10
|
+
|
|
11
|
+
- A local POSIX system with Unix-domain sockets and the same OS user running Codex and the Fleet/Pi sessions. The private policy store does not support Windows.
|
|
12
|
+
- Node.js **22.18 or newer** to run the TypeScript entrypoint directly. The package's general Node.js 18 engine requirement is not sufficient for this server.
|
|
13
|
+
- An Agent Fleet checkout or an unpacked/installed package containing `.pi/agent-fleet/scripts/fleet-codex-local-mcp.mjs` and its accompanying TypeScript runtime files. Keep the distribution intact; copying the entrypoint alone is insufficient. See [Fleet setup](agent-fleet-setup.md) for installing Fleet itself.
|
|
14
|
+
- Existing Fleet/Pi sessions publishing a local COMS registry. Observations, instruction receipts, and guarded question actions additionally require compatible receiver capabilities. A registry entry alone does not establish those capabilities.
|
|
15
|
+
- A Codex host that can launch a local STDIO MCP server and supplies its real conversation ID in each tool call's `params._meta.threadId`. This implementation requires a UUID thread ID; optional `sessionId` metadata must also be valid when supplied. Generic MCP support alone is not enough.
|
|
16
|
+
|
|
17
|
+
The server negotiates MCP protocol `2025-06-18`. Local Codex desktop use has been exercised, including a voice-to-agent handoff that performed discovery and read-policy checks. That does **not** establish support for every Codex build, another voice host, mobile access, or voice-driven mutations. Check the tools and results in your actual conversation before relying on them.
|
|
18
|
+
|
|
19
|
+
## Configure the local server
|
|
20
|
+
|
|
21
|
+
Install and maintain the server with the deterministic command described in the [Codex MCP setup guide](codex-setup.md):
|
|
22
|
+
|
|
23
|
+
```bash
|
|
24
|
+
agent-fleet set-codex-mcp status
|
|
25
|
+
agent-fleet set-codex-mcp install
|
|
26
|
+
```
|
|
27
|
+
|
|
28
|
+
It registers `af-local-read` in the user-level Codex configuration through `codex mcp add`, installs the `fleet-native-voice` skill copy under `~/.agents/skills`, and records the decision behind the user-level default. The setup guide also holds the launcher flag reference and defaults, the manual project-local alternative and its shadowing rule, migration from a project-local table, and the removal order. Those steps are not repeated here.
|
|
29
|
+
|
|
30
|
+
After installation, verify in a **new** chat with `fleet_diagnostics {}` and `fleet_discover` as the setup guide describes. Registration exposes the tools in every chat and grants no access; a missing policy namespace or an access denial before enrollment is expected and is not a reason to invent metadata or copy another conversation's state. If tools or host identity are unavailable, stop and report that limitation. A shell invocation, a legacy CLI command, or another chat's identity is not a substitute for a native tool call.
|
|
31
|
+
|
|
32
|
+
## Connect a target and manage permissions
|
|
33
|
+
|
|
34
|
+
A target is identified by `project`, `peer`, **and the exact `sessionId`/`startedAt` epoch**. Use the names returned by discovery; examples below use the illustrative project `my-project` and peer `orchestrator`. Resolve ambiguity and stale or partial discovery before granting access.
|
|
35
|
+
|
|
36
|
+
1. Discover candidates:
|
|
37
|
+
```json
|
|
38
|
+
{"project":"my-project","peer":"orchestrator"}
|
|
39
|
+
```
|
|
40
|
+
Send this to `fleet_discover`. Registry presence is not task status, responsiveness, or completion.
|
|
41
|
+
2. Prepare read enrollment with `fleet_binding_prepare`:
|
|
42
|
+
```json
|
|
43
|
+
{"intent":"enroll_read","project":"my-project","peer":"orchestrator"}
|
|
44
|
+
```
|
|
45
|
+
3. Show the returned exact target, old and proposed rights, effects, and expiry to the user. Wait for their explicit confirmation **after** that preview.
|
|
46
|
+
4. Call `fleet_binding_commit` with only the actual returned `proposalId` in the same chat. Never guess or reuse an unrelated proposal ID.
|
|
47
|
+
5. Check with `fleet_check_read`, then request `fleet_observe` when an actual peer observation is needed. A successful policy check does not itself read a transcript or probe the receiver.
|
|
48
|
+
|
|
49
|
+
Each target requires its own reviewed binding. The server chooses the binding alias; `bindingAlias` can select an existing binding, not supply an arbitrary new friendly name. Conversational labels are conveniences, not authorization.
|
|
50
|
+
|
|
51
|
+
Other `fleet_binding_prepare` intents use the same preview/confirmation/commit sequence:
|
|
52
|
+
|
|
53
|
+
| Intent | Effect |
|
|
54
|
+
| --- | --- |
|
|
55
|
+
| `rebind_read` | Select a newly discovered exact epoch explicitly. It does not carry send permission forward. |
|
|
56
|
+
| `grant_send` | Add send permission to an existing current read binding. It sends nothing by itself. |
|
|
57
|
+
| `revoke_send` | Remove send permission while keeping read. |
|
|
58
|
+
| `forget` | Remove the binding's read and send rights. It does not stop Pi or undo work already accepted. |
|
|
59
|
+
| `initialize_empty` | Create a missing, empty policy namespace for the current chat, with no target or rights. This intent accepts **no** project, peer, or alias. It is optional, not a prerequisite to read enrollment, and refuses an existing namespace. |
|
|
60
|
+
|
|
61
|
+
`read` never implies `send`. A send grant never authorizes an invented human answer. After a target restart or replacement, do not silently retarget; use fresh discovery and an explicitly confirmed rebind. Policy checks can persistently mark stale local rights as downgraded.
|
|
62
|
+
|
|
63
|
+
Proposals expire after **300 seconds** and are held in server memory. A server restart loses pending proposals, not durable rights or receipts. After expiry, state change, or restart, prepare again and obtain a new confirmation. There are shared limits of 16 pending proposals per context, 64 per process, and eight active calls.
|
|
64
|
+
|
|
65
|
+
## Observe one or more sessions
|
|
66
|
+
|
|
67
|
+
| Tool | Input and behavior |
|
|
68
|
+
| --- | --- |
|
|
69
|
+
| `fleet_discover` | Required `project`, optional `peer`; bounded registry candidates, without enrollment or peer reads. |
|
|
70
|
+
| `fleet_check_read` | `project`, `peer`; fresh registry-only exact-epoch policy check, potentially downgrading stale local rights. |
|
|
71
|
+
| `fleet_observe` | `project`, `peer`; bounded transcript, progress, and question projections from the authorized exact session. Channels can be unavailable or partial independently. |
|
|
72
|
+
| `fleet_overview` | `targets`: 1–4 explicit unique `{project,peer}` objects; separate bounded summaries and errors per target. |
|
|
73
|
+
| `fleet_resync` | Same target list; fresh observations plus durable **local** cursor/question-observation metadata updates. Reports continuity gaps and resets. |
|
|
74
|
+
| `fleet_watch` | Same target list and integer `durationSeconds` from 1–10; active bounded observation, at most six frames, with local cursor writes. |
|
|
75
|
+
|
|
76
|
+
For example, send this to `fleet_overview` or `fleet_resync`:
|
|
77
|
+
|
|
78
|
+
```json
|
|
79
|
+
{"targets":[{"project":"frontend","peer":"orchestrator"},{"project":"backend","peer":"orchestrator"}]}
|
|
80
|
+
```
|
|
81
|
+
|
|
82
|
+
These calls do not auto-enroll targets, answer questions, or write instructions to Pi. Resync and watch are not filesystem-read-only: they maintain local observation state. Watch ends; there is no background subscription, unsolicited notification, idle-model wake, or transcript mirror. Cancellation stops the observer, not Pi, and does not promise rollback of reads or local writes already started.
|
|
83
|
+
|
|
84
|
+
Treat `partial`, source timestamps, channel statuses, and gap reasons as part of the result. An unavailable question source does not mean there are no pending questions. Summary question previews are not sufficient for answering: use `fleet_observe` and the full question-action preview. Neither an idle registry entry nor a successful observation establishes whole-task completion.
|
|
85
|
+
|
|
86
|
+
## Send an instruction
|
|
87
|
+
|
|
88
|
+
Both current read and send permission are required. Sending can trigger a Pi turn; it is not a promise of immediate steering or task completion. The receiver's canonical COMS/Herdr instruction route must also be available; a successful observation does not prove send readiness.
|
|
89
|
+
|
|
90
|
+
1. Use a stable, unique `requestId` for the intended action and the exact current target. Call `fleet_instruction_prepare` with:
|
|
91
|
+
```json
|
|
92
|
+
{
|
|
93
|
+
"requestId":"instruction-001",
|
|
94
|
+
"target":{"project":"my-project","peer":"orchestrator","sessionId":"<actual sessionId>","startedAt":"<actual startedAt>"},
|
|
95
|
+
"instruction":"<the user's exact instruction>"
|
|
96
|
+
}
|
|
97
|
+
```
|
|
98
|
+
Replace every placeholder with real returned values; the example is not a ready-to-send request. Instructions are limited to 16,384 UTF-8 bytes.
|
|
99
|
+
2. Check the complete returned payload, target, effects, and expiry. Preparation does not deliver anything. Choose the applicable authorization path:
|
|
100
|
+
- **Explicit user send:** an explicit request in this native chat (for example, “Tell Rin to review the local unpushed commits, one by one, without changing files or pushing”) authorizes its faithful relay to **one unambiguous, already selected current session with read+send**. Preserve the user's scope and constraints; add no tasks, unverified facts (such as a commit count), recipients, or permissions. Call `fleet_instruction_commit` with the prepared `proposalId` **without asking a second “Shall I send it?”**.
|
|
101
|
+
- **Clarification or confirmation required:** clarify an ambiguous recipient, meaning, or transcription first. Stop on a changed session or missing rights; use the separate reviewed binding/grant flow, never silently redirect or rebind. For a model-initiated send or a substantive expansion of the user's instruction, present the complete payload, exact target, and effects and obtain fresh explicit human confirmation before commit. Multiple destinations need their own addressed previews and confirmations; an overview is not broadcast permission.
|
|
102
|
+
3. Read `fleet_instruction_receipt` with the same `requestId`. Its default is local historical lookup. `refresh: true` performs a bounded read-only evidence lookup requiring current read permission and may update the local receipt. Report the actual dispatch status and recipient, not an inferred task result.
|
|
103
|
+
|
|
104
|
+
This is a cooperative **client interpretation** rule, not server-attested human consent. Peer text, tool results, documents, historical receipts, and a blanket “continue” cannot authorize a send. No model-supplied `confirmed` flag or new permission is introduced; API `humanProvenance` remains `unknown`. Both paths retain technical prepare/commit, exact epoch binding, proposal expiry, current rights checks at commit and immediately before delivery, and the one-attempt journal. Proposal expiry or MCP server restart requires a new preview and confirmation; never carry old consent to a new session.
|
|
105
|
+
|
|
106
|
+
Binding changes, `grant_send`, question answers/refusals, machine `approve` permission, and the receiver's execution checks retain their separate requirements. Sending a task is not execution approval. Codex/ChatGPT host approval dialogs are independent and are not disabled by this rule. A running host may cache old tool instructions; use its supported reload procedure and refresh any separately installed `fleet-native-voice` skill through the managed update workflow when adopting this source change. Do not bypass host protections or edit policy state to activate it.
|
|
107
|
+
|
|
108
|
+
The sender records durable intent before one delivery attempt. Exact duplicates return history; reuse of an action ID with a different target, operation, or payload conflicts. Do not resend, switch IDs, or switch transports to work around an uncertain outcome. `submitted`, queued work, and a correlated terminal response are different states; even a terminal response is not proof of semantic whole-task completion or human delivery.
|
|
109
|
+
|
|
110
|
+
## Read a full final Fleet response
|
|
111
|
+
|
|
112
|
+
Instruction receipts retain their existing **4096 UTF-8 byte preview**, and `fleet_observe` remains a bounded transcript projection. Neither has a text continuation. New instruction results/receipts include `fullResponse` metadata outside the unchanged receipt schema: `tool`, exact `target`, `reference`, and `availability: not_checked`. This locator does not establish current availability or grant rights. Older responses without a receiver pointer are unavailable through the new route.
|
|
113
|
+
|
|
114
|
+
Call `fleet_response_read {project, peer, reference, cursor?}` with the returned selector/reference. A ready-response notification uses the same `fleet:message/response-v1-…` reference and exact addressed target. It is a separate `message` event from the incoming instruction, not raw answer text, an approval, or proof of completion. Notifications must already be explicitly enabled; reading does not enable them.
|
|
115
|
+
|
|
116
|
+
The tool makes a read-only addressed COMS RPC to the same live Fleet session. Like other observation tools, its existing policy checks may durably downgrade stale local rights, so the MCP `readOnlyHint` remains false; this does not authorize peer writes or new grants. Fleet reads the exact final assistant entry through its existing session manager. Codex does **not** open a file or choose a path. The receiver retains only bounded locator metadata and a session-history pointer, not a new answer archive; the client does not save response text into its action journal or report store.
|
|
117
|
+
|
|
118
|
+
Each page exposes `text`, UTF-8 `offsetBytes`, `nextOffsetBytes`, `totalBytes`, stable `sourceSha256`/`projectionVersion`, `chunkSha256`, `redacted`, source state/stop reason, and `nextCursor`/`eof`. Text is at most 2048 UTF-8 bytes per page with intact code points. The serialized wire frame is bounded to 16 KiB. An assistant entry is limited to 4096 content blocks. Both raw public text and its allowed projection are limited to 1 MiB; excess yields `RESPONSE_TOO_LARGE`, never a silently truncated success. Thinking and tool data are excluded. All public text blocks of the final message are joined in order with newlines, then scrubbed together before pagination, including multiline credentials and unterminated private keys.
|
|
119
|
+
|
|
120
|
+
Read from offset zero, follow every `nextCursor`, and require contiguous offsets, the same target/reference/projection/digest, and final `eof: true`, `nextCursor: null`. Tests reconstruct the bytes and verify the whole digest deterministically. Do not claim the model computed a SHA hash without an actual verifier. A server EOF alone proves neither receipt of earlier pages nor semantic task success; inspect `sourceState` separately.
|
|
121
|
+
|
|
122
|
+
A cursor survives an MCP connection/process restart **only while the same Fleet receiver/session, source and current read rights remain available**. It is bound to the native namespace and exact source version. Revocation, changed epoch/source, malformed cursor, unsupported RPC, missing pointer and `NOT_READY` remain explicit failures/gaps. Never merge versions, retarget automatically, or resend an instruction to recover missing text. Closing/restarting Fleet discards its in-memory locator index; no historical-access guarantee or archive is introduced. The index holds at most 4096 locators per live receiver and does not silently evict issued locators at capacity.
|
|
123
|
+
|
|
124
|
+
The source catalogue now contains **28 tools**. The new tool is appended, preserving existing tool order and schemas; the exact 24/27-tool notification compatibility catalogues remain distinct from current full-response support. Source tests and a separately spawned MCP process are not evidence that an existing Desktop host has adopted the new tool or that a running Fleet receiver has reloaded this handler.
|
|
125
|
+
|
|
126
|
+
## Answer or refuse a Pi question
|
|
127
|
+
|
|
128
|
+
An actual human response, current read+send rights, and a compatible guarded question receiver are required. Native Pi questions are **not** Damage Control permission dialogs or general approval requests.
|
|
129
|
+
|
|
130
|
+
1. Obtain the full pending question, options, effective flags, exact target, `questionId`, and `toolCallId`. Never select a response automatically or infer one from a general instruction to continue.
|
|
131
|
+
2. Call `fleet_answer_prepare` with `requestId`, the exact `target`, `questionId`, `toolCallId`, `operation`, and `response`:
|
|
132
|
+
- Freeform: `operation: "answer"`, `response: {"kind":"freeform","text":"<actual human answer>"}`.
|
|
133
|
+
- Selection: `operation: "answer"`, `response: {"kind":"selection","selections":["<exact option>"],"comment":"<optional human comment>"}`. Include a comment or multiple selections only when allowed by the question.
|
|
134
|
+
- Explicit refusal: `operation: "refusal"`, `response: null`.
|
|
135
|
+
3. Show the full prepared question, options/flags, normalized response, and effects. Obtain a **new** explicit human confirmation before `fleet_answer_commit {"proposalId":"<actual proposalId>"}`.
|
|
136
|
+
4. Use `fleet_answer_receipt {"requestId":"<same requestId>"}` for historical metadata, or an explicitly requested `refresh: true` for current guarded evidence when authorized.
|
|
137
|
+
|
|
138
|
+
The receiver revalidates the exact question and uses first-winner arbitration. Changed, already resolved, expired, unsupported, or incompletely addressable questions are not redirected elsewhere. A refusal resolves only that question; it is **not** Pi-turn cancellation. Selecting an option named “Stop” is a selection, not the refusal operation.
|
|
139
|
+
|
|
140
|
+
Question and response JSON are each bounded to 16,000 bytes, with up to 64 options. Guarded receiver receipts are bounded and volatile; missing evidence after a restart or a capacity limit remains unknown, not permission to retry. Receipt metadata is not an archive of answer text.
|
|
141
|
+
|
|
142
|
+
## Machine permission verification
|
|
143
|
+
|
|
144
|
+
Run `npm run test:damage-relay` for the no-network local relay regressions. It is
|
|
145
|
+
also included in `test:full`. It uses temporary stores/sockets and synthetic
|
|
146
|
+
decisions, not live approve grants or Desktop wake. The expanded checkpoint run
|
|
147
|
+
passed 360 regression tests plus 34 manifest tests; host acceptance remains open.
|
|
148
|
+
|
|
149
|
+
## Machine Damage Control permissions (source checkpoint)
|
|
150
|
+
|
|
151
|
+
This is a separate channel from human questions. `send` is **not** `approve`.
|
|
152
|
+
New/read-only bindings have no machine approval right. Use the existing reviewed
|
|
153
|
+
`fleet_binding_prepare` / `fleet_binding_commit` flow with `grant_approve` (or
|
|
154
|
+
`revoke_approve`) only after an explicit binding decision. The preview declares
|
|
155
|
+
allowlist format v2 and old-writer quiescence; the first approve grant upgrades that
|
|
156
|
+
namespace without dropping existing read/send. V1-only writers refuse v2. Never
|
|
157
|
+
roll back state or silently restore approve after rebind/epoch changes. `forget`
|
|
158
|
+
removes all rights. No live grant or host reload is performed by source tests.
|
|
159
|
+
|
|
160
|
+
- `fleet_permissions {target, operation, snapshot}`: operation is `list`, `inspect`
|
|
161
|
+
or `receipt`. List requires `snapshot:null`; inspect/receipt use the exact
|
|
162
|
+
returned snapshot. Current read authority is required before and after the RPC.
|
|
163
|
+
Separate list avoids changing the old observation wire schema.
|
|
164
|
+
- `fleet_permission_receipt {requestId}`: read this native chat's durable local
|
|
165
|
+
attempt history without a snapshot, remote refresh or restored rights. A foreign
|
|
166
|
+
chat cannot inherit it. This is the safe lookup after a lost ACK or MCP restart.
|
|
167
|
+
- `fleet_permission_decide {requestId, snapshot, decision, reason}`: after inspecting
|
|
168
|
+
the full action, Codex independently assesses its connection and necessity for
|
|
169
|
+
the exact task/revision. Decision is only `allow_once` or `deny`. A nonempty,
|
|
170
|
+
specific reason and current separate `approve` authority are required. This
|
|
171
|
+
machine operation does not ask the human to reconfirm each scoped decision.
|
|
172
|
+
It never grants turn/session exemptions or answers a human-only question.
|
|
173
|
+
|
|
174
|
+
Pi races local human and machine decisions through one latch, then revalidates
|
|
175
|
+
its action/owner/task, policy, deadline and denial epoch before issuing/consuming
|
|
176
|
+
one-use authority. Machine provenance is `codex_machine`, never human consent.
|
|
177
|
+
The host-supplied native thread UUID is recorded; the Codex turn is **not attested**
|
|
178
|
+
and is explicitly `null`. The trust boundary remains cooperative same-UID routing,
|
|
179
|
+
not cryptographic host identity. An independently approved second binding is not
|
|
180
|
+
cryptographically excluded; this is not the Desktop event-routing proof.
|
|
181
|
+
|
|
182
|
+
`<policy-root>.permissions` is an isolated durable reservation namespace using the
|
|
183
|
+
existing atomic journal engine, not a migration or rewrite of instruction/answer
|
|
184
|
+
history. Remote attempt is marked before bytes are written. Repeated IDs return
|
|
185
|
+
historical state without replay, including after MCP restart; changed payloads
|
|
186
|
+
conflict. A reserved/unknown state is not failure or permission to issue another
|
|
187
|
+
ID. Query the Pi receipt read-only when still authorized. The protocol distinguishes
|
|
188
|
+
latch acceptance from `effect:consumed` (one-use authority, **not tool completion**),
|
|
189
|
+
`not_applied`, and a legacy child `response_queued_unconfirmed`. Pi receipts are
|
|
190
|
+
bounded/volatile; Pi restart or reset can leave unknown evidence.
|
|
191
|
+
|
|
192
|
+
Baseline path asks, destructive bash asks, D6, exact triage actions and new
|
|
193
|
+
exact-action child requests are wired to the shared one-winner permission latch. Children carry a per-spawn epoch;
|
|
194
|
+
stale children, changed task/revision, clipped/unsafe action presentation and
|
|
195
|
+
`zero_access` do not acquire machine authority. Existing local human scopes remain
|
|
196
|
+
available; machine decisions stay once-only. Material triage waivers remain human-only; exact triage actions can receive
|
|
197
|
+
explicit `codex_machine` allow-once/deny under the separate approve contract. Child machine denial is explicitly
|
|
198
|
+
reported as `machine_deny`, not user refusal; expiration/cancellation/staleness do
|
|
199
|
+
not populate a human-denial cache. Child effect receipt is only response queued,
|
|
200
|
+
not proof that its command ran.
|
|
201
|
+
|
|
202
|
+
The optional user-owned `.pi/damage-control-auto-policy.yaml` is not created or
|
|
203
|
+
managed by the installer. Missing/empty/invalid means no automatic policy grant.
|
|
204
|
+
It requires a safe, user-owned non-group/world-writable file and enclosing `.pi`
|
|
205
|
+
directory; no permissions are repaired automatically. Format is `{version:1,rules:[]}`.
|
|
206
|
+
Each explicit rule has `id,decision,source,category,tool,pattern,operation,expiresAt,reason`;
|
|
207
|
+
source is `damage_control_path`, decisions `deny|allow_once`, categories
|
|
208
|
+
`zero_access|read_only|no_delete`, tools `read|write|edit|bash`, operations
|
|
209
|
+
`read|write|delete`, and expiry a canonical UTC ISO timestamp. Zero-access allow is
|
|
210
|
+
forbidden. Only one static safe operand and one exact matching rule qualify.
|
|
211
|
+
Conservative/multi-path/truncated/secret presentation never qualifies. Policy
|
|
212
|
+
digest changes invalidate pending authority; D6 holds still run and cannot be
|
|
213
|
+
bypassed. Audit/UI label this distinct source `auto_policy`. No permissive rule is
|
|
214
|
+
shipped or activated.
|
|
215
|
+
|
|
216
|
+
Local tests exercise real temporary sockets, policy stores and enforcement gates,
|
|
217
|
+
not model inference or a loaded Desktop host. Production source hooks and a durable ingress/outbox are implemented but remain
|
|
218
|
+
explicitly disabled in the live host. The core's in-memory receipts alone are not
|
|
219
|
+
the durable outbox. No live activation is implied by passing fixtures.
|
|
220
|
+
|
|
221
|
+
### Experimental opt-in Desktop queue adapter
|
|
222
|
+
|
|
223
|
+
`CodexUnixQueueRpc` and `CodexDesktopQueueHost` implement the externally
|
|
224
|
+
smoke-tested shared queue path, not a direct connection to the Desktop stdio
|
|
225
|
+
app-server. RPC connects only to an already running private Unix socket using
|
|
226
|
+
WebSocket Upgrade. The official private user-owned symlink alias is accepted only
|
|
227
|
+
for one hop to a separately checked canonical private socket; alias/target identity
|
|
228
|
+
is rechecked before initialization and addressed writes. No CLI fallback, daemon
|
|
229
|
+
start/resume, auth-file discovery, listener or automatic activation is provided.
|
|
230
|
+
|
|
231
|
+
The host must explicitly supply the exact binding/epoch, externally confirmed
|
|
232
|
+
loaded Desktop mapping, authority revision, bounded confirmation expiry, accepted
|
|
233
|
+
`user_input` transport, daemon/Desktop versions and schema hashes. Those are
|
|
234
|
+
host-supplied evidence, not cryptographic attestation or runtime binary discovery.
|
|
235
|
+
Tuple admission is exact: daemon `0.166.0` with Desktop `0.166.0-alpha.5.1`
|
|
236
|
+
or `0.166.0-alpha.6`, each requiring add-schema SHA256
|
|
237
|
+
`60f25b7d3e3357c215bef9d2fe7047200f545e030dca9e00527ac0620cd0c88f`
|
|
238
|
+
and response-schema SHA256
|
|
239
|
+
`c11b9772788427b19c9f9decf0765e2b41cbdead1aa9c3d39f42fdd70b339314`.
|
|
240
|
+
The alpha.6 addition has measured schema equality and offline unit/mocked admission
|
|
241
|
+
coverage only; it does not extend historical live smoke results or authorize M1.
|
|
242
|
+
For the exact Desktop `0.167.0-alpha.3.2` tuple the daemon pins above still
|
|
243
|
+
apply, and approval also requires `desktopAddSchema`
|
|
244
|
+
`f6466a8f06cd4e68ec1999d06a0d70a1ddc339c0cb1baf12ee966a999fead17c`
|
|
245
|
+
and `desktopResponseSchema`
|
|
246
|
+
`896a82716b432a8aa4d190591bcfa86adacdacb91ec5ec2c9c75ff6562561256`.
|
|
247
|
+
The measured schema delta adds an audio input variant. Fleet queue traffic
|
|
248
|
+
remains text only. This tuple has offline coverage and awaits live idle-wake
|
|
249
|
+
verification. Other versions are refused even if schemas match. `hostInstance` remains an
|
|
250
|
+
operator-supplied binding identifier, not a discovered runtime credential.
|
|
251
|
+
The daemon must report the target `notLoaded`; a known
|
|
252
|
+
daemon-loaded target is refused because that consumer might win instead. This
|
|
253
|
+
read cannot exclude a concurrent consumer or prove that Desktop remains loaded.
|
|
254
|
+
Do not refresh mapping confirmation merely because saved thread metadata exists.
|
|
255
|
+
|
|
256
|
+
Events carry only validated kind-specific `fleet:<kind>/...` references and a
|
|
257
|
+
fixed `fleet_machine` envelope, never raw commands/questions. They are visibly
|
|
258
|
+
machine-origin despite arriving as native **user** input; `humanApproval:false`
|
|
259
|
+
never supplies human consent or `approve`. Exact `fleet:probe/<eventId>` message
|
|
260
|
+
references request only `FLEET_PROBE_ACK <eventId>`, without tools or permissions.
|
|
261
|
+
|
|
262
|
+
Question references use one shared lossless codec across ingress, bridge, packet and
|
|
263
|
+
reply. Existing safe ASCII addresses remain byte-for-byte
|
|
264
|
+
`fleet:question/<questionId>/<toolCallId>` (preserving old journal digests).
|
|
265
|
+
Other already-valid opaque tool-call IDs use
|
|
266
|
+
`fleet:question:v2/<questionId>/<base64url(UTF-16LE(toolCallId))>`. This includes
|
|
267
|
+
provider composite IDs containing `|`, separators, Unicode and lone JS surrogates;
|
|
268
|
+
no trimming, Unicode normalization, truncation or replacement is allowed. The
|
|
269
|
+
native limit is unchanged: 1–256 UTF-16 code units, excluding C0 and DEL controls.
|
|
270
|
+
Decoders reject padding, noncanonical base64url, odd byte lengths, extra components,
|
|
271
|
+
invalid IDs and v2 aliases of legacy-safe IDs. Only canonical v2 questions receive
|
|
272
|
+
the derived maximum reference length of **798** characters (96-character question
|
|
273
|
+
ID plus at most 512 encoded bytes). Other references retain the 512-character
|
|
274
|
+
bound. References remain data, never commands or approval.
|
|
275
|
+
|
|
276
|
+
The private `<stateRoot>.desktop-events` journal reserves an attempt before the
|
|
277
|
+
queue write. `queue_accepted` and `submissionId` mean server ACK, not model
|
|
278
|
+
observation/execution; cached ACKs carry `historical:true`. Lost ACKs remain
|
|
279
|
+
`unknown` across adapter restarts and are never resent. Read-only reconciliation
|
|
280
|
+
can establish presence in queue, never model delivery from an empty queue. The
|
|
281
|
+
backend client message ID is correlation, not an assumed idempotency guarantee.
|
|
282
|
+
Use the same persistent state root on restart; this journal does not persist
|
|
283
|
+
unaccepted producer events. Queued expired events are not automatically recalled;
|
|
284
|
+
Pi still validates the original permission deadline and action independently.
|
|
285
|
+
|
|
286
|
+
Remaining budgets of 10 seconds or less are unsupported. External marker artifacts
|
|
287
|
+
corroborate ~7.1s emission-to-input and ~9.3s emission-to-reply for one idle sample
|
|
288
|
+
(using the externally reported emission timestamp), and ~61.6s/~63.4s when queued
|
|
289
|
+
behind a busy turn containing an intentional `sleep 60`. Dispatch after the
|
|
290
|
+
previous turn ended was **55ms**; the ~61.6s is total wait, not pure channel latency.
|
|
291
|
+
Isolation in both directions was covered by that second external smoke. Neither a
|
|
292
|
+
5s SLA nor general deadline reliability is established; queue wait remains inside
|
|
293
|
+
the original deadline.
|
|
294
|
+
These limited host samples do not establish installed adapter compatibility,
|
|
295
|
+
restart safety of the whole bridge, or real model permission decisions. No live
|
|
296
|
+
adapter message or grant is activated by the local tests.
|
|
297
|
+
|
|
298
|
+
A separate explicitly authorized **single** temporary adapter probe on 2026-11-04
|
|
299
|
+
passed: one queue/add, ACK at +26ms, correlated input in the exact control thread
|
|
300
|
+
at +1.418s, exact machine reply at +8.005s inside the original 30s deadline, and
|
|
301
|
+
zero model tool calls. No daemon startup/resume, real Fleet grant or persistent
|
|
302
|
+
bridge activation occurred. This establishes that concrete adapter/host round-trip,
|
|
303
|
+
not independent UI rendering, exclusive process identity, a 5s SLA or whole-bridge
|
|
304
|
+
restart/permission-decision acceptance. SDK saved-thread reads,
|
|
305
|
+
active-turn first-party delivery, MCP watch, CLI exit success and Voice alone
|
|
306
|
+
remain insufficient idle-wake proof.
|
|
307
|
+
|
|
308
|
+
### Explicit production event ingress (disabled until approved)
|
|
309
|
+
|
|
310
|
+
Real producers publish only exact owner/source IDs, kind, original time/deadline
|
|
311
|
+
and content digests: COMS before prompt ACK/dispatch; questions after their reply
|
|
312
|
+
offer is installed; permissions before returning the hold. Question refs contain
|
|
313
|
+
both question ID and toolCallId, so a bounded observation page is not mistaken for
|
|
314
|
+
the complete set. `fleet_permissions` additionally accepts an optional list
|
|
315
|
+
`cursor`; follow `nextCursor`, use inspect for the full safe action. Independent
|
|
316
|
+
`permissionState` is visible in observe/overview/resync/watch. Missing capability
|
|
317
|
+
is `unsupported`, not zero pending. Existing v1 request shapes remain valid.
|
|
318
|
+
|
|
319
|
+
`/af-codex-events prepare /absolute/private/config.json` reads a user-owned,
|
|
320
|
+
regular, no-follow private file and produces readiness only: no socket contact,
|
|
321
|
+
journal write, daemon lifecycle or grant. `enable` is a **separate explicit user
|
|
322
|
+
activation**, not a recommendation to execute it during source verification;
|
|
323
|
+
`disable` stops the in-process consumer and aborts in-flight binding work.
|
|
324
|
+
No configuration is discovered or created by the installer. The bootstrap
|
|
325
|
+
rechecks that the private config is unchanged before ingress/host write. Changing
|
|
326
|
+
or removing it revokes notification work; it never auto-applies new authority.
|
|
327
|
+
Session reset also stops the runtime. Explicit re-prepare/enable is needed for a
|
|
328
|
+
new mapping; existing read/send/approve rights are not rewritten by this lifecycle.
|
|
329
|
+
|
|
330
|
+
Required config keys are exactly `optedIn`, `stateRoot`, `socketPath`, `approval`,
|
|
331
|
+
`loadedCatalog`, `messageBudgetMs`. `approval` is the exact typed Desktop mapping
|
|
332
|
+
(binding id/epoch/hostInstance/threadId/fleetOwner, target epoch, authority revision,
|
|
333
|
+
confirmation expiry, accepted user-input provenance, mapping evidence and pinned
|
|
334
|
+
versions/schemas). Message notifications with no source expiry use an explicitly
|
|
335
|
+
chosen 10001–60000ms notification budget; question/permission deadlines are never
|
|
336
|
+
extended. The question's own lifetime remains independently authoritative.
|
|
337
|
+
|
|
338
|
+
`loadedCatalog` is a normalized **actual** native diagnostics observation (tool
|
|
339
|
+
names, identity assurance, Node and protocol), not a copied source tool list.
|
|
340
|
+
The earlier observation of **21**, `host_context_validated`, Node 25.2.1 and
|
|
341
|
+
protocol 2025-06-18 is historical. On 2026-11-04 the human relayed new direct main
|
|
342
|
+
observations: **24/24**, completed receiver reload/rebind, and permissionState
|
|
343
|
+
available/complete/zero pending. The old main catalog blocker is closed; do not
|
|
344
|
+
request the same restart/reload again. Preserve/normalize the actual host result,
|
|
345
|
+
not the source list. Actual per-chat native identity, current mapping, rights and
|
|
346
|
+
journal formats still need their own evidence; the catalog count grants none.
|
|
347
|
+
`assessLoadedFleetCatalog` is a pure comparator; it never reloads.
|
|
348
|
+
|
|
349
|
+
`<stateRoot>.event-ingress` durably stores a versioned payload before source ACK,
|
|
350
|
+
with binding/authority revision, stable identity and digest conflict checks. It
|
|
351
|
+
uses the existing synchronous locked/fsync/rename/marker engine, isolated from
|
|
352
|
+
instruction/answer/permission stores. A synchronous validated ingress ACK is
|
|
353
|
+
mandatory; a Promise or void sink cannot claim durability. Startup validates the
|
|
354
|
+
whole journal and reconciles before accepting new source traffic. Only committed
|
|
355
|
+
**never-attempted** events may be re-admitted; attempted/unknown records call
|
|
356
|
+
`reconcile()` read-only, even when the queue is empty or a remote reservation is
|
|
357
|
+
absent. Attempt is marked before calling the host, conservatively before actual
|
|
358
|
+
bytes. Duplicate ingested and terminal/history states are distinct. Partial
|
|
359
|
+
unattempted payload slots are operator-visible and may only be filled by the
|
|
360
|
+
matching original source; attempted unknowns are never blindly repaired/resubmitted.
|
|
361
|
+
|
|
362
|
+
The state root must stay stable across restart. Capacity is bounded (256 intents
|
|
363
|
+
per namespace); unsafe/full/corrupt/changed state fails closed rather than rotating
|
|
364
|
+
or forgetting replay fences. With an explicitly active consumer, durable admission
|
|
365
|
+
failure refuses the source admission; disabled notifications retain prior source
|
|
366
|
+
behavior. Exactly one consumer is allowed for an owner: no cross-chat broadcast.
|
|
367
|
+
No production consumer, config, grant or daemon has been activated by this work.
|
|
368
|
+
|
|
369
|
+
#### Safe rollout and notification checks after a source fix
|
|
370
|
+
|
|
371
|
+
Source/unit-test success is not proof the running consumer loaded the new codec.
|
|
372
|
+
Arrange a separately approved reload at a safe work boundary, deploying the ingress,
|
|
373
|
+
bridge, Desktop packet adapter and reply reader together before emitting v2. Retain
|
|
374
|
+
both notification journals and the stable state root; do not rewrite, repair or
|
|
375
|
+
replay attempted/unknown events. Legacy-safe references remain readable and unchanged.
|
|
376
|
+
Recheck the actual current AF owner/epoch, first-party current Codex identity,
|
|
377
|
+
explicit Desktop mapping and expiry, existing rights and loaded native catalog.
|
|
378
|
+
A replaced AF session requires the separate reviewed binding/permission flow, not
|
|
379
|
+
implicit transfer. Run the supported `prepare` on the exact private configuration
|
|
380
|
+
and present its actual result; any `enable`, new expiry or configuration change
|
|
381
|
+
requires its own explicit approval. Do not renew from saved thread metadata alone.
|
|
382
|
+
|
|
383
|
+
For a live question check, use **one new** harmless human question only after the
|
|
384
|
+
updated components and current mapping are established. Record the actual question ID,
|
|
385
|
+
raw toolCallId and versioned reference; do not retry an old failed/unknown question
|
|
386
|
+
or force a simpler provider ID. Distinguish source durable admission, queue ACK,
|
|
387
|
+
arrival in the exact Codex chat and model ACK of the same event ID. A model ACK is
|
|
388
|
+
not a human answer: no automatic `fleet_answer_commit` or access decision. The
|
|
389
|
+
human may answer/cancel the test through the ordinary reviewed question workflow.
|
|
390
|
+
|
|
391
|
+
For Damage Control, inspect an existing or naturally arising legitimate pending
|
|
392
|
+
request through the current native `fleet_permissions` list/inspect interface and
|
|
393
|
+
correlate its actual event ID/reference. Never touch a protected file, replay a
|
|
394
|
+
blocked command or fabricate a live permission request to provoke an alert. Do not
|
|
395
|
+
allow, deny or cancel real work merely to complete a notification test. With no
|
|
396
|
+
legitimate request, only an isolated mock-producer test is possible under this scope.
|
|
397
|
+
Preserve the original deadline: remaining budgets of 10 seconds or less are
|
|
398
|
+
unsupported, not a reason to extend a hold, change IDs or resend an unknown event.
|
|
399
|
+
|
|
400
|
+
The direct dependency remains pinned **ws@8.22.0**, already locally available. It
|
|
401
|
+
provides the tested Unix connection hook, early frame limits, UTF-8/fragment/close
|
|
402
|
+
handling and disabled redirects. The built-in browser-compatible WebSocket API
|
|
403
|
+
does not expose these same direct `socketPath/createConnection/maxPayload` controls;
|
|
404
|
+
an alternative dispatcher/transport would need its own equivalence/security proof.
|
|
405
|
+
This is a dependency choice, not a blanket vulnerability or host attestation.
|
|
406
|
+
|
|
407
|
+
### Reply routing and prepared latency tests
|
|
408
|
+
|
|
409
|
+
`replyFleetEvent` checks the trusted current binding **and independent actual
|
|
410
|
+
native thread identity**, not fields copied from an event. A permission reply
|
|
411
|
+
lists/pages, fully inspects and rechecks the pending snapshot, then calls the
|
|
412
|
+
separate machine decision API. Every native API still checks its own rights and
|
|
413
|
+
freshness. No unknown tool result triggers an automatic retry. A normal message
|
|
414
|
+
produces a machine summary, never an implicit instruction/command.
|
|
415
|
+
|
|
416
|
+
Questions retain `fleet_answer_prepare/commit` and their human-confirmation
|
|
417
|
+
contract. The default reply path prepares only. Commit needs a separate genuine
|
|
418
|
+
host-controlled human interaction; a model-supplied boolean/notification is not
|
|
419
|
+
consent. Synthetic callback tests are not evidence of a real human. Permission
|
|
420
|
+
machine decisions do not acquire this human provenance, and human-only waivers
|
|
421
|
+
are not routed through the question API.
|
|
422
|
+
|
|
423
|
+
`RelayLatencyTrace`/`runLatencySeries` provide 20–200-sample drivers and metadata-only
|
|
424
|
+
phase records. Original source deadlines and monotonic local times are retained;
|
|
425
|
+
remote clock error must be explicit. Recorded host input/history is not model
|
|
426
|
+
awareness; ACK is not Pi applied. Pi receipt timings include decision and actual
|
|
427
|
+
one-use consume timestamps, not command completion; legacy child response-queued
|
|
428
|
+
has no fabricated applied timestamp. Local series are `fixture_no_model`.
|
|
429
|
+
Live series need a new coordinator approval callback for the exact profile/count.
|
|
430
|
+
All source remains uncommitted; commit/reload/grants/live runs remain user decisions.
|
|
431
|
+
|
|
432
|
+
## Local diagnostics, snapshots, and legacy history
|
|
433
|
+
|
|
434
|
+
These operations do not launch a shell, contact another host, or deliver a Pi instruction:
|
|
435
|
+
|
|
436
|
+
| Tool | Behavior |
|
|
437
|
+
| --- | --- |
|
|
438
|
+
| `fleet_diagnostics {}` | Validates native host context and reports the Node runtime, protocol, available tool names/effects, and whether an archive root is configured. It does not probe registry, sockets, or peer capabilities. Use `fleet_check_read` and an authorized observation for those separate checks. |
|
|
439
|
+
| `fleet_receipts {offset?,limit?}` | Local current-chat instruction/question action history, including retained unknown results. Permission history uses `fleet_permission_receipt` separately. Offset is zero-based; limit is 1–8 (default 8). Follow `nextOffset` to enumerate. It does not refresh or replay actions. |
|
|
440
|
+
| `fleet_snapshot {project,peer}` | Explicitly observe one currently read-authorized exact target and retain its redacted report locally. Unlike ordinary observation, this requests content retention for later replay. It makes no peer write. |
|
|
441
|
+
| `fleet_replay {project,peer}` | Return the last explicitly saved snapshot, marked `historical_only`, with its original target and observation time. It performs no live permission/source check and remains local historical data after revocation. It is not evidence about the current target or a restored grant. |
|
|
442
|
+
| `fleet_legacy_history {section,offset?,limit?,sourceSha256?}` | Read the configured original CLI archive for this exact native conversation UUID. Section is `selection` or `runtime`; no model-selected paths, IDs, identities, roots, or grants are accepted. |
|
|
443
|
+
|
|
444
|
+
Snapshots are private files under `<policy-root>.reports`. Each saved report is bounded to 96 KiB; saving a new snapshot for the same chat/selector replaces its prior snapshot under an exclusive lock. There is no automatic snapshot on ordinary observe/resync. Revocation does not erase already retained historical data. Preserve snapshots on a capacity, unsafe-path, or uncertain-write error; do not interpret a replay as fresh progress.
|
|
445
|
+
|
|
446
|
+
### Preserve existing CLI data without replaying it
|
|
447
|
+
|
|
448
|
+
If you have legacy data, configure the optional launcher pair `"--legacy-state-root", "/absolute/path/to/original/fleet-client"` after reviewing that archive location yourself. It must be a normalized, private, real directory separate from registry and native state roots. The reader does not discover `$CODEX_HOME` or read environment identity. Do not enable a second writer or mutate the archive during transfer. In-flight legacy writer locks block new native actions when this archive is configured.
|
|
449
|
+
|
|
450
|
+
The reader selects only `<actual-native-thread-uuid>.json` and its `.runtime` sibling. A detached old selection may be absent while its runtime history remains accessible. Original files, timestamps, digests, wire IDs, and unknown outcomes are not rewritten or imported into native action records. Stored CLI selection never creates native read/send rights. A legacy command or question-response ID reserves that ID against a new native send/answer when the archive is configured; a reserved ID must be inspected as history, not retried under another ID. Configure the archive **before** issuing potentially overlapping new actions. Omitting the option does not establish that legacy delivery history is empty.
|
|
451
|
+
|
|
452
|
+
Example: `fleet_legacy_history {"section":"runtime","offset":0,"limit":8192}` returns a redacted JSON fragment, `sourceSha256` for the original bytes, and `nextOffset`. For every continuation (`offset > 0`), pass the first page's `sourceSha256` along with `nextOffset`. Missing hashes are rejected; if the original changes or disappears, `HISTORY_STATE_CHANGED` stops pagination. Discard incomplete fragments and, if still needed, explicitly restart from offset zero. Never combine versions. Follow offsets until null and reconstruct the full text before interpreting a record. Offsets count string characters, not bytes. Inputs are bounded to 4 MiB per original file, 131,072 JSON nodes, and 8,192 characters per fragment. Sensitive keys are redacted; individual text values longer than 4,096 characters are wholly withheld rather than incompletely scrubbed. Raw originals remain intact. This is bounded historical access, not an assurance that arbitrary credentials have been identified or an immutable proof of remote delivery.
|
|
453
|
+
|
|
454
|
+
There is no automatic mapping between different native/legacy conversation IDs, profiles, or hosts. Keep those original archives untouched; cross-identity transfer needs a separate explicit mapping decision. Hosts lacking the required native MCP context are not supported by the replacement workflow. Archive access does not enable such a host, authenticate a human, or grant another conversation access.
|
|
455
|
+
|
|
456
|
+
## Persistent state and journal upgrades
|
|
457
|
+
|
|
458
|
+
Native state is partitioned by configured host profile, principal, and host-supplied thread ID. It is separate from the legacy CLI. Other conversations do not inherit this chat's grants.
|
|
459
|
+
|
|
460
|
+
- `--policy-root`: bindings and revision-bound permissions.
|
|
461
|
+
- `<policy-root>.observations`: bounded local cursors and question-observation metadata, not a transcript archive or proof that a human saw an event.
|
|
462
|
+
- `<policy-root>.reports`: explicitly retained, redacted historical snapshots; separate from observation cursors and deduplication.
|
|
463
|
+
- `<policy-root>.permissions`: isolated machine-permission attempt reservations; no raw action text, no automatic replay and no tool-completion claim.
|
|
464
|
+
- `<policy-root>.instructions`: a shared action journal for instruction, answer, and refusal IDs. Historical receipts remain readable after revocation; remote refresh still requires current read rights.
|
|
465
|
+
|
|
466
|
+
Existing private state directories must be owned by the current user with mode `0700`; state files must be private, owned regular files with mode `0600`. Do not reuse an arbitrary shared directory or manually edit state files to bypass a refusal. Missing state is normally created through the reviewed binding workflow above, not by the installer.
|
|
467
|
+
|
|
468
|
+
The shared instruction/answer action journal is bounded to 256 total records and 8 MiB. There is no automatic purge, deduplication eviction, reset, or corrupt-state repair. Instruction bodies and full question/answer text are not archived there, but bounded instruction results can be retained. This is not a promise to remove content from Codex, chat logs, tool logs, or other systems.
|
|
469
|
+
|
|
470
|
+
An existing **native v1 action journal** is history-only until explicitly upgraded. It is not the legacy CLI journal. `INTENT_FORMAT_UPGRADE_REQUIRED` blocks new action writes and remote receipt refresh on that format.
|
|
471
|
+
|
|
472
|
+
To upgrade, stop old journal writers and in-flight mutating actions first. Call `fleet_journal_upgrade_prepare {}` in the actual native chat. Review the returned namespace, snapshot hash/revision, exact backup location, expiry, and rollback boundary. The operator must confirm writer quiescence; the server does not infer it from an idle pane or file lock. Obtain a fresh human confirmation, then call `fleet_journal_upgrade_commit` with that proposal's ID.
|
|
473
|
+
|
|
474
|
+
The upgrade preserves original receipts and makes an immutable raw backup before the same-file v2 transition. It grants no rights and makes no Pi call. A validated existing v2 journal is reported without another migration proposal. An uncertain commit must be reconciled, not replayed blindly. Old v1 binaries reject v2; safe rollback keeps a v2-capable reader/data or remains offline. **Never restore an old journal snapshot as a retry mechanism.**
|
|
475
|
+
|
|
476
|
+
## Troubleshooting
|
|
477
|
+
|
|
478
|
+
| Symptom or code | What to check or do |
|
|
479
|
+
| --- | --- |
|
|
480
|
+
| Missing server/tools; `LOCAL_MCP_START_FAILED` | Run `agent-fleet set-codex-mcp status` ([setup guide](codex-setup.md#troubleshooting)): it reports the Node version, launcher paths, entry state, a project shadow, and refusals without dumping private state; `update` repairs the launcher after a Node change. For a manual project-local table, check the Node version, absolute executable/entrypoint paths, all four launcher flags, the intact runtime distribution, project trust, and any `enabled_tools` filter. Use only the host's supported user-controlled reload/reopen procedure. |
|
|
481
|
+
| `IDENTITY_REQUIRED`, `CONFLICTING_HOST_CONTEXT` | The host did not provide accepted per-call metadata or supplied conflicting context. Stop; do not fabricate a UUID, set `CODEX_THREAD_ID` as a substitute, or borrow another chat's identity. |
|
|
482
|
+
| `REGISTRY_MISSING`, stale/partial discovery, ambiguous candidates | Check the actual registry root and intended running project's COMS registration. Resolve ambiguity before binding; do not invent entries or select by working directory alone. |
|
|
483
|
+
| Partial discovery with `ENTRY_INVALID`, especially immediately after startup | A record is missing or has invalid required metadata; this code alone does not identify the field. Updated COMS publishes `heartbeat_at` on its first atomic registration, avoiding the old pre-heartbeat gap. Finite measured context above 100% is valid metadata, not an admission limit; do not clamp it or fabricate an entry. Preserve safe diagnostics and verify the loaded receiver version. Any user-controlled reload changes the epoch and requires explicit rebind; never reload or restore send rights automatically. |
|
|
484
|
+
| `STATE_MISSING` | This namespace has no usable policy state. For a genuinely new namespace, use reviewed enrollment or optional `initialize_empty`. If state was previously present, investigate missing data rather than resetting it. |
|
|
485
|
+
| `NOT_ALLOWED`, `REBIND_REQUIRED`, stale target | Check this chat's exact target and rights. Read does not imply send. A new epoch requires explicit rebind and a separate send grant if needed. Never grant automatically to clear an error. |
|
|
486
|
+
| Expired/missing proposal; `STATE_CHANGED`, `BUSY` | Discard stale previews. Once conflicting work finishes, prepare a new proposal and ask for a new confirmation. Do not bypass revision checks or delete locks. |
|
|
487
|
+
| `RPC_UNSUPPORTED`, `SOURCE_NOT_CONFIGURED`, `INSTRUCTION_UNSUPPORTED`, `ANSWER_UNSUPPORTED` | The registered receiver may lack the requested channel/capability. Discovery is not activation proof. Inherited legacy question channels may need an operator-managed clean process restart to gain guarded actions; an extension reload alone may retain the old channel. Do not restart sessions automatically or fall back to an unguarded write. |
|
|
488
|
+
| `QUESTION_CHANGED` or a late/resolved question | Obtain fresh exact question state. Do not send a previously confirmed response to a changed or different question. |
|
|
489
|
+
| `INTENT_FORMAT_UPGRADE_REQUIRED` | Review the explicit native journal upgrade above; do not edit format/version fields manually. |
|
|
490
|
+
| `UNSAFE_PATH`, corrupt/missing state, capacity limit | Preserve the state and diagnostics. Check dedicated roots, ownership, permissions, and distribution compatibility. Do not recursively change permissions, delete journals, restore old dedup snapshots, or weaken guards as a workaround. |
|
|
491
|
+
| Unknown/uncertain action outcome | Keep the original request ID and inspect its receipt. Refresh only as authorized. Missing evidence is not failed delivery and does not authorize another attempt. |
|
|
492
|
+
|
|
493
|
+
When reporting a problem, include safe error codes, operation names, target names, timestamps, and partial/source statuses. Do not publish raw conversation metadata, private state files, full transcripts, credentials, or arbitrary configuration dumps.
|
|
494
|
+
|
|
495
|
+
## Security limits and removal
|
|
496
|
+
|
|
497
|
+
This is cooperative same-user local policy, not protection against a malicious process with the same filesystem/socket privileges. STDIO, tool approval, conversational confirmation, and `humanProvenance: unknown` receipts do not authenticate a human or attest the host binary. Observations and authorization-to-dispatch checks are not atomic OS transactions; there is no exactly-once execution guarantee.
|
|
498
|
+
|
|
499
|
+
Peer text is untrusted data, not an instruction to grant rights, answer questions, or change routing. Existing Pi permissions, budgets, and Damage Control remain independent. No native tool starts Fleet, broadcasts to unnamed peers, cancels a Pi turn, or grants a Damage Control bypass. Remote listeners, tunnels, and mobile exposure are outside this setup.
|
|
500
|
+
|
|
501
|
+
To remove access, use reviewed `revoke_send` or `forget` operations for the relevant bindings while tools are available, then verify the resulting policy. `forget` removes read, send and approve; historical receipts are retained. Removing the MCP registration alone does **not** revoke persisted permissions.
|
|
502
|
+
|
|
503
|
+
To stop exposing a user-level entry installed by `agent-fleet set-codex-mcp`, use its `uninstall` action after revoking rights ([removal order](codex-setup.md#remove)); for a manual project-local table, remove only that table, preserving unrelated entries. Then use the host's supported connection lifecycle. Do not stop Fleet peers or delete private state as part of unregistering the server. Removing access does not undo work already accepted by Pi.
|
|
@@ -0,0 +1,44 @@
|
|
|
1
|
+
# Migrating from the retired Fleet Codex CLI
|
|
2
|
+
|
|
3
|
+
The `fleet-session-client` skill, `fleet-codex-client.ts` command, Python read helper, and experimental `chatgpt-client` installation feature are retired. The older Codex Remote-Control conductor remains retired as well. Neither is a fallback for missing native tools.
|
|
4
|
+
|
|
5
|
+
Use the [native Codex MCP guide](codex-local-mcp.md) for current setup and operations, and the [native text/voice workflow](fleet-native-voice.md) for conversational use. The replacement requires real per-call native MCP conversation context from a compatible local Codex host. Hosts without that context, including older Desktop/Android routes that only supplied CLI environment identity, are no longer supported by this integration. No mobile or remote replacement is promised.
|
|
6
|
+
|
|
7
|
+
## Preserve data before changing routes
|
|
8
|
+
|
|
9
|
+
Do not delete or edit the original CLI selection files, `.runtime` journals, receipts, locks, or backups. A successful instruction, an unknown outcome, and an unavailable result are different records; none authorizes replay. Existing files remain historical evidence even when their writer is retired.
|
|
10
|
+
|
|
11
|
+
The CLI default was `$CODEX_HOME/fleet-client`, or the `fleet-client` directory under the default Codex home when `CODEX_HOME` was absent. A custom `--state-dir` could choose another location. Determine your actual original location yourself; the native server does not search the environment or import arbitrary directories.
|
|
12
|
+
|
|
13
|
+
For the same real conversation UUID, explicitly configure `--legacy-state-root` on the native launcher and use `fleet_legacy_history` to read redacted, paginated historical data. See [local history and archive access](codex-local-mcp.md#local-diagnostics-snapshots-and-legacy-history) for the exact limits and identity boundary. Original wire IDs, digests, timestamps, and unknown results are not transformed into new actions. A detached selection can be absent while its runtime journal remains readable. Old action IDs remain reserved against new native delivery when the archive is configured; configure it before overlapping new instructions or answers.
|
|
14
|
+
|
|
15
|
+
This is read-only compatibility, not automatic transfer of grants or a rewrite into the native action ledger. Different conversation IDs require an explicit mapping decision outside this automatic path; their archives must remain intact. A root or state validation error must not be worked around by resetting, copying another chat's files, or inventing identity.
|
|
16
|
+
|
|
17
|
+
## Replace the operations, not the old command
|
|
18
|
+
|
|
19
|
+
| Retired CLI use | Native replacement |
|
|
20
|
+
| --- | --- |
|
|
21
|
+
| sessions / select / detach | `fleet_discover`; reviewed binding enroll/rebind/forget with exact target/epoch and fresh confirmation |
|
|
22
|
+
| status | `fleet_check_read`; registry/policy evidence, not task completion |
|
|
23
|
+
| summary / activity / questions | `fleet_observe`; full bounded question data for addressing a question |
|
|
24
|
+
| resync | `fleet_resync`, with explicit local cursor writes and gaps |
|
|
25
|
+
| watch | Explicit bounded `fleet_watch` while the conversation is active; longer requested periods use bounded calls within that time budget, not a daemon |
|
|
26
|
+
| send | `fleet_instruction_prepare`, one commit, exact receipt; an explicit user send to one unambiguous selected current session with read+send needs no second confirmation; otherwise clarify or obtain preview confirmation as described in the [native send contract](codex-local-mcp.md#send-an-instruction) |
|
|
27
|
+
| answer / cancel-question | `fleet_answer_prepare`, actual human answer or explicit question-scoped refusal, fresh preview confirmation, one commit and receipt; not Pi-turn cancellation |
|
|
28
|
+
| replay | Explicit `fleet_snapshot` followed by local historical `fleet_replay`; legacy saved reports are available through the configured archive reader |
|
|
29
|
+
| receipts | `fleet_receipts`, exact instruction/answer receipt lookup, and opt-in `fleet_legacy_history` for old records |
|
|
30
|
+
| doctor / describe / help | `fleet_diagnostics`, actual MCP tool schemas, registry/read checks and authorized capability observations; static diagnostics are not receiver readiness proof |
|
|
31
|
+
|
|
32
|
+
The native interface changes authorization deliberately: legacy selection does not grant read or send, read does not grant send, and an old send grant does not supply a human answer. Keep technical prepare/commit for every send. Explicit user sends follow the limited no-second-confirmation path above; binding changes, answers, model-initiated sends and substantively expanded instructions retain their own reviewed confirmation requirements. Do not send an old uncertain action again to demonstrate migration.
|
|
33
|
+
|
|
34
|
+
## Repair retired installer selections
|
|
35
|
+
|
|
36
|
+
Requests for `--features chatgpt-client` or `--features codex-remote` are rejected without enabling another integration. If either name remains in `.ai/agent-fleet.json`, even as `false`, use the normal installer config-repair flow to review the exact removals and explicitly approve its backup. `--yes` alone is not config-repair approval. Keep the workspace's other feature selections.
|
|
37
|
+
|
|
38
|
+
Config repair removes only the retired desired-state settings; it does not install native MCP, enroll targets, copy history, remove private journals, or stop sessions. Review the normal installer plan before removing previously owned runtime/skill files. Modified or foreign files retain the installer's ownership/conflict protections. Old release snapshots remain for migration and are not an active source of the retired integration.
|
|
39
|
+
|
|
40
|
+
Install/configure the native server and skill separately using the linked guides. Removing an old skill or registration does not revoke existing native permissions, and no upgrade should silently rebind a target or change the running Pi session.
|
|
41
|
+
|
|
42
|
+
## Limits and evidence
|
|
43
|
+
|
|
44
|
+
Native operations and archive compatibility have local fixture, subprocess, and temporary-socket regression coverage. These are not proof that your host has loaded the new source or that all hosts work. The bounded real desktop voice route established native discovery/read-policy checks, not every mutation through voice. Verify actual native availability in your intended conversation without fabricating context, granting automatically, restarting Fleet, or restoring a retired transport.
|