@chankov/agent-fleet 1.0.1 → 1.0.2
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/extensions/package-lock.json +77 -45
- package/.pi/harnesses/agent-hub/README.md +109 -18
- package/.pi/harnesses/agent-hub/budget-continuation.ts +73 -0
- package/.pi/harnesses/agent-hub/capability-confirmation.ts +47 -0
- package/.pi/harnesses/agent-hub/capability-packs.ts +180 -0
- package/.pi/harnesses/agent-hub/context-budget-snapshot.ts +105 -0
- package/.pi/harnesses/agent-hub/context-pressure.ts +191 -0
- package/.pi/harnesses/agent-hub/hub-state-audit.js +18 -0
- package/.pi/harnesses/agent-hub/index.ts +988 -422
- package/.pi/harnesses/agent-hub/package.json +1 -1
- package/.pi/harnesses/agent-hub/posture.ts +72 -17
- package/.pi/harnesses/agent-hub/run-budget.js +12 -13
- package/.pi/harnesses/agent-hub/spawn.ts +14 -2
- package/.pi/harnesses/ask-user-remote/index.ts +59 -4
- package/.pi/harnesses/coms/package.json +1 -1
- package/.pi/harnesses/damage-control-continue/package.json +1 -1
- package/.pi/harnesses/lib/context-budget-child-prompt.ts +179 -0
- package/.pi/harnesses/lib/context-budget-hub-prompt.ts +116 -0
- package/.pi/harnesses/lib/context-budget-view.ts +107 -0
- package/.pi/harnesses/lib/context-budget.ts +114 -0
- package/.pi/harnesses/lib/fleet-detail-view.ts +92 -2
- package/.pi/harnesses/package-lock.json +73 -80
- package/.versions/1.0.2/.pi/agents/dispatch-policy.yaml +32 -0
- package/.versions/1.0.2/.pi/agents/model-profiles.yaml +27 -0
- package/.versions/1.0.2/.pi/agents/peers.yaml +142 -0
- package/.versions/1.0.2/.pi/agents/teams.yaml +41 -0
- package/.versions/1.0.2/.pi/damage-control-rules.yaml +297 -0
- package/.versions/1.0.2/.pi/extensions/agent-fleet-update-check/README.md +57 -0
- package/.versions/1.0.2/.pi/extensions/agent-fleet-update-check/index.ts +161 -0
- package/.versions/1.0.2/.pi/extensions/agent-fleet-update-check/package.json +6 -0
- package/.versions/1.0.2/.pi/extensions/btw/README.md +120 -0
- package/.versions/1.0.2/.pi/extensions/btw/index.ts +914 -0
- package/.versions/1.0.2/.pi/extensions/btw/package.json +6 -0
- package/.versions/1.0.2/.pi/extensions/btw/state.js +179 -0
- package/.versions/1.0.2/.pi/extensions/chrome-devtools-mcp/README.md +55 -0
- package/.versions/1.0.2/.pi/extensions/chrome-devtools-mcp/index.ts +93 -0
- package/.versions/1.0.2/.pi/extensions/chrome-devtools-mcp/package.json +6 -0
- package/.versions/1.0.2/.pi/extensions/compact-and-continue/README.md +42 -0
- package/.versions/1.0.2/.pi/extensions/compact-and-continue/index.ts +120 -0
- package/.versions/1.0.2/.pi/extensions/compact-and-continue/package.json +6 -0
- package/.versions/1.0.2/.pi/extensions/mcp-bridge/README.md +46 -0
- package/.versions/1.0.2/.pi/extensions/mcp-bridge/index.ts +206 -0
- package/.versions/1.0.2/.pi/extensions/mcp-bridge/package.json +6 -0
- package/.versions/1.0.2/.pi/extensions/package-lock.json +1175 -0
- package/.versions/1.0.2/.pi/extensions/package.json +9 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/README.md +205 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/config.ts +198 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/examples/stt.json +35 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/index.ts +248 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/package.json +6 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/providers.ts +214 -0
- package/.versions/1.0.2/.pi/extensions/pi-voice-stt/recorder.ts +158 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/README.md +1156 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/access-approval.ts +135 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/artifacts.js +146 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/assertion-ledger.js +58 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/backend-policy.js +174 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/budget-continuation.ts +73 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/capability-confirmation.ts +47 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/capability-packs.ts +180 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/context-budget-snapshot.ts +105 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/context-pressure.ts +191 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/context-window.js +106 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/delegate.ts +435 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/docs-lane.js +122 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/drift-watchdog.js +211 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/evidence-rules.js +96 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/external-blocker.js +85 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/footer.ts +35 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/helpers.ts +266 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/hub-state-audit.js +73 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/index.ts +9909 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-control.ts +9 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-invoke-journal.ts +131 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-invoke.ts +92 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-lifecycle.ts +13 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-publisher.ts +59 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-recovery.ts +3 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-runtime.ts +3 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/monitor-session-bridge.ts +266 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/package.json +7 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/peer-spawn-plan.ts +116 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/posture.ts +158 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/provider-semaphore.js +138 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/research-retention.js +34 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/research-watchdog.ts +40 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/return-contract.js +293 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/return-extract.js +75 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/review-findings.js +102 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/run-budget.js +539 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/run-namespace.js +87 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/safety-routing.ts +22 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/scope-gate.js +103 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/session-health.js +110 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/spawn.ts +366 -0
- package/.versions/1.0.2/.pi/harnesses/agent-hub/version.ts +42 -0
- package/.versions/1.0.2/.pi/harnesses/ask-user-remote/README.md +53 -0
- package/.versions/1.0.2/.pi/harnesses/ask-user-remote/index.ts +532 -0
- package/.versions/1.0.2/.pi/harnesses/ask-user-remote/race-core.js +106 -0
- package/.versions/1.0.2/.pi/harnesses/coms/README.md +100 -0
- package/.versions/1.0.2/.pi/harnesses/coms/index.ts +1767 -0
- package/.versions/1.0.2/.pi/harnesses/coms/package.json +7 -0
- package/.versions/1.0.2/.pi/harnesses/coms/version.ts +42 -0
- package/.versions/1.0.2/.pi/harnesses/damage-control-continue/README.md +124 -0
- package/.versions/1.0.2/.pi/harnesses/damage-control-continue/index.ts +536 -0
- package/.versions/1.0.2/.pi/harnesses/damage-control-continue/package.json +7 -0
- package/.versions/1.0.2/.pi/harnesses/damage-control-continue/version.ts +42 -0
- package/.versions/1.0.2/.pi/harnesses/lib/coms-registry-entry.ts +94 -0
- package/.versions/1.0.2/.pi/harnesses/lib/context-budget-child-prompt.ts +179 -0
- package/.versions/1.0.2/.pi/harnesses/lib/context-budget-hub-prompt.ts +116 -0
- package/.versions/1.0.2/.pi/harnesses/lib/context-budget-view.ts +107 -0
- package/.versions/1.0.2/.pi/harnesses/lib/context-budget.ts +114 -0
- package/.versions/1.0.2/.pi/harnesses/lib/damage-control-shared.ts +197 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fixtures/context-budget-profiles.json +11 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fixtures/hub-prompt-operator.legacy.txt +73 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fixtures/hub-prompt-orchestrator.legacy.txt +73 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-dashboard-ops.ts +155 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-dashboard-view.ts +140 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-detail-view.ts +133 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-overlay.ts +26 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-panel.ts +34 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-read-model.ts +131 -0
- package/.versions/1.0.2/.pi/harnesses/lib/fleet-selection.ts +24 -0
- package/.versions/1.0.2/.pi/harnesses/lib/herdr-client.ts +472 -0
- package/.versions/1.0.2/.pi/harnesses/lib/herdr-presence.ts +399 -0
- package/.versions/1.0.2/.pi/harnesses/lib/hermes-monitor-events.ts +136 -0
- package/.versions/1.0.2/.pi/harnesses/lib/hermes-monitor-herdr.ts +28 -0
- package/.versions/1.0.2/.pi/harnesses/lib/hermes-monitor-model.ts +18 -0
- package/.versions/1.0.2/.pi/harnesses/lib/hermes-monitor-registry.ts +6 -0
- package/.versions/1.0.2/.pi/harnesses/lib/hermes-monitor-socket.ts +163 -0
- package/.versions/1.0.2/.pi/harnesses/lib/hermes-monitor-store.ts +31 -0
- package/.versions/1.0.2/.pi/harnesses/lib/spawned-peers.js +190 -0
- package/.versions/1.0.2/.pi/harnesses/package-lock.json +2039 -0
- package/.versions/1.0.2/.pi/harnesses/package.json +10 -0
- package/.versions/1.0.2/.pi/prompts/af-build.md +24 -0
- package/.versions/1.0.2/.pi/prompts/af-code-simplify.md +22 -0
- package/.versions/1.0.2/.pi/prompts/af-plan.md +16 -0
- package/.versions/1.0.2/.pi/prompts/af-review.md +16 -0
- package/.versions/1.0.2/.pi/prompts/af-set-hermes-telegram.md +15 -0
- package/.versions/1.0.2/.pi/prompts/af-set-hermes-watchdog.md +6 -0
- package/.versions/1.0.2/.pi/prompts/af-ship.md +17 -0
- package/.versions/1.0.2/.pi/prompts/af-spec.md +15 -0
- package/.versions/1.0.2/.pi/prompts/af-test.md +19 -0
- package/.versions/1.0.2/.pi/skills/bowser/SKILL.md +145 -0
- package/.versions/1.0.2/.version +1 -0
- package/.versions/1.0.2/agents/architect.md +24 -0
- package/.versions/1.0.2/agents/bowser.md +27 -0
- package/.versions/1.0.2/agents/builder.md +54 -0
- package/.versions/1.0.2/agents/code-reviewer.md +208 -0
- package/.versions/1.0.2/agents/deep-researcher.md +32 -0
- package/.versions/1.0.2/agents/documenter.md +18 -0
- package/.versions/1.0.2/agents/orchestrator.md +20 -0
- package/.versions/1.0.2/agents/plan-reviewer.md +83 -0
- package/.versions/1.0.2/agents/planner.md +70 -0
- package/.versions/1.0.2/agents/releaser.md +24 -0
- package/.versions/1.0.2/agents/researcher.md +28 -0
- package/.versions/1.0.2/agents/security-auditor.md +167 -0
- package/.versions/1.0.2/agents/test-engineer.md +153 -0
- package/.versions/1.0.2/agents/web-debugger.md +43 -0
- package/.versions/1.0.2/agents/web-performance-auditor.md +193 -0
- package/.versions/1.0.2/bin/catalog/harness-runtime-closure.json +38 -0
- package/.versions/1.0.2/codex/CONDUCTOR.md +148 -0
- package/.versions/1.0.2/docs/ARCHITECTURE.md +369 -0
- package/.versions/1.0.2/docs/agent-fleet-setup.md +468 -0
- package/.versions/1.0.2/docs/claude-code-coms-bridge.md +128 -0
- package/.versions/1.0.2/docs/codex-remote-conductor.md +170 -0
- package/.versions/1.0.2/docs/coms-hermes-bridge.md +189 -0
- package/.versions/1.0.2/docs/getting-started.md +98 -0
- package/.versions/1.0.2/docs/npm-install.md +292 -0
- package/.versions/1.0.2/docs/pi-extensions.md +439 -0
- package/.versions/1.0.2/docs/pi-setup.md +74 -0
- package/.versions/1.0.2/docs/skill-anatomy.md +150 -0
- package/.versions/1.0.2/hermes/README.md +347 -0
- package/.versions/1.0.2/hermes/desktop-plugins/agent-fleet-herdr/plugin.js +1291 -0
- package/.versions/1.0.2/hermes/desktop-plugins/agent-fleet-herdr/presentation.js +672 -0
- package/.versions/1.0.2/hermes/desktop-plugins/agent-fleet-monitor/plugin.js +19 -0
- package/.versions/1.0.2/hermes/desktop-plugins/agent-fleet-monitor/state.js +10 -0
- package/.versions/1.0.2/hermes/gates/gate-o-capability.md +8 -0
- package/.versions/1.0.2/hermes/gates/gate-o-validator-fixtures.json +1 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/activity.py +770 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/coms_registry.py +134 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/coms_send.py +350 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/herdr_source.py +193 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/manifest.json +4 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/plugin_api.py +444 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/sessions.py +205 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/tasks.py +300 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/dashboard/watch.py +819 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-herdr/plugin.yaml +8 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-monitor/dashboard/adapter.py +139 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-monitor/dashboard/manifest.json +11 -0
- package/.versions/1.0.2/hermes/plugins/agent-fleet-monitor/dashboard/plugin_api.py +50 -0
- package/.versions/1.0.2/hermes/skills/hub-conductor/SKILL.md +84 -0
- package/.versions/1.0.2/hermes/skills/hub-liaison/SKILL.md +84 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/SKILL.md +76 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/manifest.json +1 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog.py +249 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_actions.py +124 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_contract.py +76 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_delivery.py +10 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_engine.py +128 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_judgment.py +41 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_recovery.py +103 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_state.py +163 -0
- package/.versions/1.0.2/hermes/skills/hub-watchdog/scripts/watchdog_transport.py +138 -0
- package/.versions/1.0.2/hermes/watchdog-tests/__init__.py +0 -0
- package/.versions/1.0.2/hooks/coms-stop-hook.mjs +84 -0
- package/.versions/1.0.2/install-manifest.json +2745 -0
- package/.versions/1.0.2/justfile +272 -0
- package/.versions/1.0.2/manifest-meta.json +882 -0
- package/.versions/1.0.2/references/accessibility-checklist.md +160 -0
- package/.versions/1.0.2/references/definition-of-done.md +67 -0
- package/.versions/1.0.2/references/fleet-coordination-patterns.md +151 -0
- package/.versions/1.0.2/references/observability-checklist.md +91 -0
- package/.versions/1.0.2/references/orchestration-patterns.md +370 -0
- package/.versions/1.0.2/references/performance-checklist.md +153 -0
- package/.versions/1.0.2/references/prompting-patterns.md +380 -0
- package/.versions/1.0.2/references/security-checklist.md +179 -0
- package/.versions/1.0.2/references/testing-patterns.md +236 -0
- package/.versions/1.0.2/scripts/codex-conductor.ts +41 -0
- package/.versions/1.0.2/scripts/codex-remote-control.ts +167 -0
- package/.versions/1.0.2/scripts/coms-claude-bridge.ts +411 -0
- package/.versions/1.0.2/scripts/coms-cli.ts +484 -0
- package/.versions/1.0.2/scripts/coms-hermes-bridge.ts +413 -0
- package/.versions/1.0.2/scripts/fleet.ts +185 -0
- package/.versions/1.0.2/scripts/hermes-monitor-capability-runner.ts +25 -0
- package/.versions/1.0.2/scripts/hermes-monitor-scenario.ts +474 -0
- package/.versions/1.0.2/scripts/hermes-watchdog-capability-probe.ts +45 -0
- package/.versions/1.0.2/scripts/install-hermes-plugin.sh +180 -0
- package/.versions/1.0.2/scripts/lib/claude-bridge-core.ts +210 -0
- package/.versions/1.0.2/scripts/lib/codex-conductor.ts +88 -0
- package/.versions/1.0.2/scripts/lib/codex-remote-control.ts +480 -0
- package/.versions/1.0.2/scripts/lib/coms-envelope.ts +582 -0
- package/.versions/1.0.2/scripts/lib/fake-herdr-server.ts +88 -0
- package/.versions/1.0.2/scripts/lib/fleet-command.ts +266 -0
- package/.versions/1.0.2/scripts/lib/herdr-layout.ts +281 -0
- package/.versions/1.0.2/scripts/lib/hermes-bridge-core.ts +241 -0
- package/.versions/1.0.2/scripts/lib/hermes-monitor-events.ts +1 -0
- package/.versions/1.0.2/scripts/lib/hermes-monitor-herdr.ts +2 -0
- package/.versions/1.0.2/scripts/lib/hermes-monitor-model.ts +3 -0
- package/.versions/1.0.2/scripts/lib/hermes-monitor-registry.ts +2 -0
- package/.versions/1.0.2/scripts/lib/hermes-monitor-socket.ts +2 -0
- package/.versions/1.0.2/scripts/lib/hermes-monitor-store.ts +2 -0
- package/.versions/1.0.2/scripts/lib/monitor-env.ts +146 -0
- package/.versions/1.0.2/scripts/lib/peer-launch.ts +318 -0
- package/.versions/1.0.2/scripts/lib/persona-banner.ts +62 -0
- package/.versions/1.0.2/scripts/lib/spawn-stagger.ts +101 -0
- package/.versions/1.0.2/scripts/lib/team-project.ts +213 -0
- package/.versions/1.0.2/scripts/lib/team-snapshot.ts +147 -0
- package/.versions/1.0.2/scripts/peer-banner.ts +40 -0
- package/.versions/1.0.2/scripts/peer-launch.ts +259 -0
- package/.versions/1.0.2/scripts/team-snapshot.ts +237 -0
- package/.versions/1.0.2/scripts/team-up.ts +406 -0
- package/.versions/1.0.2/skills/_internal/grilling.md +11 -0
- package/.versions/1.0.2/skills/browser-testing-with-devtools/SKILL.md +351 -0
- package/.versions/1.0.2/skills/code-review-and-quality/SKILL.md +381 -0
- package/.versions/1.0.2/skills/compound-learning/SKILL.md +137 -0
- package/.versions/1.0.2/skills/context-engineering/SKILL.md +291 -0
- package/.versions/1.0.2/skills/deprecation-and-migration/SKILL.md +206 -0
- package/.versions/1.0.2/skills/designing-agents/SKILL.md +394 -0
- package/.versions/1.0.2/skills/designing-agents/pi-harness-authoring.md +222 -0
- package/.versions/1.0.2/skills/frontend-ui-engineering/SKILL.md +328 -0
- package/.versions/1.0.2/skills/git-workflow-and-versioning/SKILL.md +316 -0
- package/.versions/1.0.2/skills/idea-refine/SKILL.md +182 -0
- package/.versions/1.0.2/skills/idea-refine/examples.md +238 -0
- package/.versions/1.0.2/skills/idea-refine/frameworks.md +99 -0
- package/.versions/1.0.2/skills/idea-refine/refinement-criteria.md +113 -0
- package/.versions/1.0.2/skills/idea-refine/scripts/idea-refine.sh +15 -0
- package/.versions/1.0.2/skills/incremental-implementation/SKILL.md +287 -0
- package/.versions/1.0.2/skills/interview-me/SKILL.md +226 -0
- package/.versions/1.0.2/skills/orchestration-verification/SKILL.md +181 -0
- package/.versions/1.0.2/skills/peer-coms/SKILL.md +90 -0
- package/.versions/1.0.2/skills/performance-optimization/SKILL.md +350 -0
- package/.versions/1.0.2/skills/planning-and-task-breakdown/SKILL.md +254 -0
- package/.versions/1.0.2/skills/security-and-hardening/SKILL.md +461 -0
- package/.versions/1.0.2/skills/spec-driven-development/SKILL.md +257 -0
- package/.versions/1.0.2/skills/using-agent-skills/SKILL.md +199 -0
- package/.versions/1.0.2/systemd/user/agent-fleet-codex-remote-control.service.in +23 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/LICENSE +21 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/README.md +11 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/api-and-interface-design/SKILL.md +294 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/browser-testing-with-devtools/SKILL.md +317 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/ci-cd-and-automation/SKILL.md +390 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/code-review-and-quality/SKILL.md +396 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/code-simplification/SKILL.md +331 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/context-engineering/SKILL.md +289 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/debugging-and-error-recovery/SKILL.md +300 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/deprecation-and-migration/SKILL.md +247 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/documentation-and-adrs/SKILL.md +278 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/doubt-driven-development/SKILL.md +243 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/frontend-ui-engineering/SKILL.md +328 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/git-workflow-and-versioning/SKILL.md +355 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/idea-refine/SKILL.md +178 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/idea-refine/examples.md +238 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/idea-refine/frameworks.md +99 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/idea-refine/refinement-criteria.md +113 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/idea-refine/scripts/idea-refine.sh +15 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/incremental-implementation/SKILL.md +249 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/interview-me/SKILL.md +225 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/observability-and-instrumentation/SKILL.md +203 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/performance-optimization/SKILL.md +350 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/planning-and-task-breakdown/SKILL.md +234 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/security-and-hardening/SKILL.md +467 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/shipping-and-launch/SKILL.md +310 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/source-driven-development/SKILL.md +194 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/spec-driven-development/SKILL.md +206 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/test-driven-development/SKILL.md +383 -0
- package/.versions/1.0.2/vendor/agent-skills-upstream/skills/using-agent-skills/SKILL.md +191 -0
- package/CHANGELOG.md +9 -0
- package/README.md +4 -2
- package/agents/orchestrator.md +4 -56
- package/docs/ARCHITECTURE.md +48 -16
- package/docs/agent-fleet-setup.md +33 -2
- package/docs/pi-extensions.md +29 -2
- package/install-manifest.json +2 -2
- package/package.json +2 -2
|
@@ -0,0 +1,369 @@
|
|
|
1
|
+
# Agent Fleet Architecture
|
|
2
|
+
|
|
3
|
+
Agent Fleet is a Pi-centered multi-agent orchestration system. This page maps
|
|
4
|
+
the runtime responsibilities and where each module lives in the repository.
|
|
5
|
+
|
|
6
|
+
## Runtime layers
|
|
7
|
+
|
|
8
|
+
| Layer | Role | Implementation |
|
|
9
|
+
| --- | --- | --- |
|
|
10
|
+
| **Pi Coding Agent** | Primary local runtime — runs the dispatcher and specialist subagents | `.pi/harnesses/`, `.pi/extensions/`, `.pi/agents/`, `.pi/prompts/` |
|
|
11
|
+
| **agent-hub** | Thin-context multi-agent harness: dispatcher + specialists + research helpers + Verification Contract | `.pi/harnesses/agent-hub/` |
|
|
12
|
+
| **Herdr** | Fleet/workspace control plane — spawns peer teams as tiled workspaces, presence via push events, snapshot/resume | [herdr.dev](https://herdr.dev); client in `.pi/harnesses/lib/herdr-client.ts`, layout in `scripts/lib/herdr-layout.ts` |
|
|
13
|
+
| **coms** | Peer communication protocol/data plane — envelope-based P2P messaging between agents | `.pi/harnesses/coms/`, `scripts/lib/coms-envelope.ts`, `scripts/coms-cli.ts` |
|
|
14
|
+
| **Claude Code bridge** | Makes an interactive Claude Code pane a bidirectional coms peer | `scripts/coms-claude-bridge.ts`, `hooks/coms-stop-hook.mjs`, `skills/peer-coms/` — see [claude-code-coms-bridge.md](claude-code-coms-bridge.md) |
|
|
15
|
+
| **Hermes bridge** | Remote human control — relays hub questions to Telegram, races phone vs. local answers, conductor/liaison skills | `scripts/coms-hermes-bridge.ts`, `.pi/harnesses/ask-user-remote/`, `hermes/skills/` — see [coms-hermes-bridge.md](coms-hermes-bridge.md) |
|
|
16
|
+
| **Hermes local monitor transport** | Local, authenticated monitor contract for Hub-owned task generations; consumers supply their own presentation | `.pi/harnesses/agent-hub/monitor-*.ts`, `.pi/harnesses/lib/hermes-monitor-{model,store,registry,socket}.ts` (with compatibility re-exports under `scripts/lib/`) — see [Hermes artifacts](../hermes/README.md#local-agent-hub-monitor-integration) and [watchdog limits](hermes-watchdog-supervisor.md) |
|
|
17
|
+
| **Hermes Desktop plugin** (`agent-fleet-herdr`) | Fleet observability surface — read-only panel of every live session joined from the coms registry, herdr presence, agent transcripts, and the monitor transport; `focus` and subagent `cancel` are its only write doors | `hermes/desktop-plugins/agent-fleet-herdr/` (Electron pane), `hermes/plugins/agent-fleet-herdr/dashboard/` (FastAPI backend), installed by `scripts/install-hermes-plugin.sh` — see [hermes-desktop-plugins.md](hermes-desktop-plugins.md) |
|
|
18
|
+
| **Codex remote-control conductor** | Experimental outbound-only, user-systemd-managed Android conductor; verified on Codex CLI 0.144.x | `scripts/codex-remote-control.ts`, `scripts/codex-conductor.ts`, `codex/CONDUCTOR.md`, `systemd/user/`; runtime under `~/.local/state/agent-fleet/codex-conductor/` — see [codex-remote-conductor.md](codex-remote-conductor.md) |
|
|
19
|
+
| **Skill library** | Lifecycle workflows and quality gates every agent follows | `skills/` (native) + `vendor/agent-skills-upstream/skills/` (vendored) — see [UPSTREAM-SKILLS.md](UPSTREAM-SKILLS.md) |
|
|
20
|
+
| **Personas** | Reusable specialist definitions, installed verbatim | `agents/`, `bin/lib/personas.js` |
|
|
21
|
+
|
|
22
|
+
## One Hub runtime, independent axes
|
|
23
|
+
|
|
24
|
+
Every public Pi launch through `just fleet` loads Fleet Core and the `agent-hub`
|
|
25
|
+
harness. Four independent choices shape that runtime:
|
|
26
|
+
|
|
27
|
+
| Axis | Choices | What it controls |
|
|
28
|
+
| --- | --- | --- |
|
|
29
|
+
| **Posture** | `operator`, `orchestrator` | Whether the main agent may use direct coding tools. Operator retains the captured `read`/`bash`/`edit`/`write` and approved extension surface; orchestrator removes direct coding tools and keeps dispatch, research, assertions, `ask_user`, and available coms/Herdr tools. |
|
|
30
|
+
| **Hub mode** | `fast`, `standard`, `strict` | Dispatch/research budgets and Verification Contract rigor. `/af-hub-mode` never grants or revokes coding tools. |
|
|
31
|
+
| **Native roster** | empty or one entry from `.pi/agents/teams.yaml` | Which in-process Pi specialist personas `dispatch_agent` may start. Roster changes do not change posture in a live session. |
|
|
32
|
+
| **Peer topology** | current terminal, Hub-only Herdr workspace, or Hub plus a `.pi/agents/peers.yaml` preset | Which separate, addressable Pi/Claude processes occupy sibling panes. Starting or spawning peers does not change posture or the native roster. |
|
|
33
|
+
|
|
34
|
+
Bare `just fleet` therefore means operator posture, an empty native roster, and
|
|
35
|
+
the current terminal. At startup an explicit `--posture` wins; otherwise
|
|
36
|
+
`--agents <roster>` implies orchestrator; otherwise operator is used. An
|
|
37
|
+
orchestrator startup requires a roster. `/af-posture` switches only posture in
|
|
38
|
+
the live session, while `/af-agents-*` changes only the native roster.
|
|
39
|
+
|
|
40
|
+
All Hub-owned slash commands remain registered in both postures. Capability packs are resolved automatically from explicit task intent, posture, tier, pending work, and compaction state—there is no activation command. Runtime **readiness** (a coms or Herdr connection) is not model-visible capability: ready-but-unrequested peer/workspace packs remain inactive. Ambiguous fleet, peer, and workspace requests are provisionally visible but require one `ask_user` confirmation before their first side effect; a rejection removes the provisional pack. Task packs persist across follow-ups and reset only through `set_task_tier(new_task: true)`, while mandatory posture and pending-operation leases remain.
|
|
41
|
+
|
|
42
|
+
`/af-context` is a standalone, read-only full-screen budget diagnostic; it separates stable prompt cost, volatile state, active schemas, and zero-cost inactive/loaded-excluded inputs. Provider totals and cache read/write fields are authoritative where Pi supplies them; it never fabricates provider usage or combines capacity percentages across Hub, child, and peer planes. Managed specialists and research children use explicit replacement prompts/context manifests with selected persona, policy, and skill paths instead of inherited global skills/context files. Runtime gates control whether an action can proceed: without coms, Herdr, a visible target, or an active roster, the corresponding command refuses with an actionable message rather than disappearing. `--browser` and `--all-extensions` expand the captured operator surface only when those optional extensions are installed.
|
|
43
|
+
|
|
44
|
+
### Context pressure and recovery
|
|
45
|
+
|
|
46
|
+
Standing prompt ceilings constrain the repeatable replacement prompt and active schemas; they do not
|
|
47
|
+
bound conversation messages or results accumulated inside a long tool loop. The Hub therefore has a
|
|
48
|
+
second, runtime pressure guard. It samples provider usage, projects a finalized tool result at
|
|
49
|
+
`message_end`, warns and exposes transient compaction support at **80%**, then at **90%** aborts the
|
|
50
|
+
pre-model continuation and waits for `agent_settled` before requesting one Pi compaction. This
|
|
51
|
+
ordering ensures the result is persisted before summarization and prevents an ordinary provider call
|
|
52
|
+
from escaping between pressure detection and recovery.
|
|
53
|
+
|
|
54
|
+
Recovery is single-flight. High-context startup or concurrent input is retained in memory and
|
|
55
|
+
replayed exactly once after success; failure retains it for `/compact` or a larger-context model.
|
|
56
|
+
`/af-context` and the status line report phase, usage, thresholds, episode, and last outcome. Session
|
|
57
|
+
entries store only numeric/enumerated pressure metadata, never prompts, tool payloads, credentials,
|
|
58
|
+
error bodies, or summaries.
|
|
59
|
+
|
|
60
|
+
A persisted native roster is also metadata, not copied configuration: only its team name is stored
|
|
61
|
+
and it is re-resolved against current teams and persona definitions. A stale or absent roster in a
|
|
62
|
+
resumed orchestrator session fails closed—posture remains orchestrator, direct tools remain removed,
|
|
63
|
+
and model input is blocked until `/af-agents-team`, an explicit `--agent-team <name>` restart, or an
|
|
64
|
+
explicit operator selection resolves it. Explicit CLI posture and roster selections retain
|
|
65
|
+
precedence.
|
|
66
|
+
|
|
67
|
+
Pi's session JSONL is the authoritative append-only record for rebuilding a session, including
|
|
68
|
+
compaction and Hub metadata. Pressure and roster recovery operate through Pi/Hub lifecycle APIs;
|
|
69
|
+
they never rewrite prior entries. Operators must not edit, truncate, reorder, or synthesize JSONL to
|
|
70
|
+
recover a session.
|
|
71
|
+
|
|
72
|
+
The project selected by `--project <name>` is the coms namespace for the Hub
|
|
73
|
+
and every standing or dynamically spawned peer. The Hub-facing peer spawn API
|
|
74
|
+
cannot override it. Native subagents remain children of that Hub and inherit
|
|
75
|
+
its repository context; peers are separate processes in their own panes.
|
|
76
|
+
|
|
77
|
+
At dispatch time, `backend: "auto"` applies
|
|
78
|
+
`.pi/agents/dispatch-policy.yaml`; `backend: "native"` bypasses same-name peer
|
|
79
|
+
substitution; and `backend: "coms"` requires a visible same-name peer and never
|
|
80
|
+
falls back to a native child. `.pi/agents/peers.yaml` owns how a peer runs
|
|
81
|
+
(`runner`, persona, model, extensions, and declared `env_file`), not whether a
|
|
82
|
+
dispatch uses it.
|
|
83
|
+
|
|
84
|
+
## Fleet hierarchy
|
|
85
|
+
|
|
86
|
+
Agent Fleet is layered on purpose. Work flows **down** (delegate); evidence and status flow **up** — as compact structured returns, never raw dumps.
|
|
87
|
+
|
|
88
|
+
```mermaid
|
|
89
|
+
flowchart TD
|
|
90
|
+
You(["You · Hermes inbound relay · Codex outbound conductor on your phone"])
|
|
91
|
+
|
|
92
|
+
subgraph HUBL["HUB — guarded runtime (bare just fleet)"]
|
|
93
|
+
Hub["agent-hub harness<br/>operator or orchestrator posture · routes tasks<br/>owns the Verification Contract on disk · keeps research dumps out of its own context"]
|
|
94
|
+
end
|
|
95
|
+
|
|
96
|
+
subgraph TEAML["OPTIONAL NATIVE ROSTER (.pi/agents/teams.yaml)"]
|
|
97
|
+
Team["default: builder · test-engineer · code-reviewer · documenter<br/>also: plan · debug · frontend · security · hotfix · release · info"]
|
|
98
|
+
end
|
|
99
|
+
|
|
100
|
+
subgraph RESL["RESEARCH HELPERS — read-only, always available"]
|
|
101
|
+
Research["researcher (fast tier — simple reads)<br/>deep-researcher (deep tier — hard, cross-cutting questions)"]
|
|
102
|
+
end
|
|
103
|
+
|
|
104
|
+
subgraph SUBL["SUB-AGENTS — focused children, narrow tools + models"]
|
|
105
|
+
Subs["planner → scout · rules · risk<br/>plan-reviewer → feasibility · deps<br/>builder → recon · verifier<br/>test-engineer → coverage-scout · conventions<br/>code-reviewer → preflight · quality · perf · docs<br/>security-auditor → recon · input-sweep · secrets-sweep"]
|
|
106
|
+
end
|
|
107
|
+
|
|
108
|
+
You -->|task| Hub
|
|
109
|
+
Hub -->|"dispatch_agent — one persona = one specialist session"| Team
|
|
110
|
+
Hub -->|"spawn_research"| Research
|
|
111
|
+
Research -.->|"findings written to disk — hub gets paths, not dumps"| Hub
|
|
112
|
+
Team -->|"subagents: block (delegate_depth ≥ 1)"| Subs
|
|
113
|
+
Subs -.->|"results return to the parent agent only"| Team
|
|
114
|
+
Team -.->|"structured return + evidence"| Hub
|
|
115
|
+
Hub -.->|"one status line: Assertions: 2✓ 1○ 1✗"| You
|
|
116
|
+
```
|
|
117
|
+
|
|
118
|
+
Every specialist session is one persona from [`agents/`](../agents/) — *skills* tell each agent **how** to work; *personas* define **who** they are (see [agents.md](agents.md)).
|
|
119
|
+
|
|
120
|
+
The same idea as a tree:
|
|
121
|
+
|
|
122
|
+
```text
|
|
123
|
+
hub (operator by default; orchestrator when selected)
|
|
124
|
+
├── optional native roster (empty by default; example: default)
|
|
125
|
+
│ ├── builder → recon · verifier
|
|
126
|
+
│ ├── test-engineer → coverage-scout · conventions
|
|
127
|
+
│ ├── code-reviewer → preflight · quality · perf · docs
|
|
128
|
+
│ └── documenter
|
|
129
|
+
├── research helpers (spawn_research, any time)
|
|
130
|
+
│ ├── researcher
|
|
131
|
+
│ └── deep-researcher
|
|
132
|
+
└── optional fleet peers (herdr + coms)
|
|
133
|
+
├── architect / releaser / web-debugger panes
|
|
134
|
+
├── Claude Code peer (coms bridge)
|
|
135
|
+
├── Hermes (phone human · inbound ask_user)
|
|
136
|
+
└── Codex Remote Control (Android · outbound coms delegation)
|
|
137
|
+
```
|
|
138
|
+
|
|
139
|
+
Composition rule: **the hub (or a slash command) orchestrates; personas do not invoke other personas as peers.** Specialists may only fan out to their configured **sub-agents**. Research helpers write findings to disk; the hub resumes specialists with paths, not raw dumps.
|
|
140
|
+
|
|
141
|
+
### Research search supervision
|
|
142
|
+
|
|
143
|
+
The shared `spawnPiAgent` seam supervises every `read`/`grep`/`find`/`ls` call from native
|
|
144
|
+
research helpers and nested delegate children. The supervisor tracks JSONL `toolCallId` values
|
|
145
|
+
independently, with a default 120-second deadline (`recon-search-timeout-s: 1..3600|off` under
|
|
146
|
+
`## agent-hub`). It is a per-tool watchdog; the whole-run bound is separate — the execution
|
|
147
|
+
mode's per-run deadline (`agent-turn-timeout-s`), which terminates a hung run as `turn_timeout`.
|
|
148
|
+
On timeout or caller cancellation it owns and terminates the child's process group (SIGTERM,
|
|
149
|
+
then SIGKILL after a bounded grace), has a separate settlement timer for missing `close`/pipe
|
|
150
|
+
drain, and reports timeout separately from cancellation. Research helpers and nested delegates
|
|
151
|
+
are each given safe process-group ownership; delegates forward parent termination so no detached
|
|
152
|
+
child is orphaned. Full pattern catalog: [references/orchestration-patterns.md](../references/orchestration-patterns.md).
|
|
153
|
+
|
|
154
|
+
### Execution modes & turn budgets
|
|
155
|
+
|
|
156
|
+
The hub enforces per-user-turn budgets in code (`run-budget.js`): `fast`/`standard`/`strict`
|
|
157
|
+
modes cap `dispatch_agent` calls, `spawn_research` calls, and active time per turn, set the
|
|
158
|
+
per-run deadline above, and control nested delegation. Exhausted budgets make the dispatch
|
|
159
|
+
tools refuse and request one Yes/No `ask_user` confirmation; Yes renews the turn in the same
|
|
160
|
+
tool loop. A normal new user message also opens a fresh turn window.
|
|
161
|
+
Specialist context pressure is measured over input + cacheRead + cacheWrite against **that
|
|
162
|
+
agent's own** model window, resolved from pi's model registry with the source recorded
|
|
163
|
+
(`context-window.js`) — measuring a 49k local model against the dispatcher's window is what
|
|
164
|
+
made readings like "315%" unactionable; anything over 100% now emits a one-time diagnostic
|
|
165
|
+
naming the window and where it came from. Specialist sessions are recycled (fresh spawn
|
|
166
|
+
instead of `-c` resume) after `session-recycle-runs` runs, at ≥60% measured context, and
|
|
167
|
+
unconditionally at a full window; a resumed session whose *projected* prompt would overflow
|
|
168
|
+
is recycled before the spawn rather than after the run. Requests to one provider are capped
|
|
169
|
+
per process (`provider-semaphore.js`: 2 in flight for `custom/*` by default, unlimited
|
|
170
|
+
elsewhere, `AGENT_HUB_PROVIDER_LIMITS` to override) — the cap is per level of the delegation
|
|
171
|
+
tree, and a nested spawn reuses its parent's permit so it can never wait on its own ancestor.
|
|
172
|
+
Configured under `## agent-hub` (`mode`, `max-dispatches-per-turn`,
|
|
173
|
+
`max-research-per-turn`, `turn-wall-time-s`, `agent-turn-timeout-s`, `session-recycle-runs`,
|
|
174
|
+
`run-history-keep`); switched live with `/af-hub-mode`.
|
|
175
|
+
|
|
176
|
+
A per-message allowance cannot bound a task, so a second envelope sits above the turn one:
|
|
177
|
+
the **task budget** (`run-budget.js`, `3×` the turn envelope) counts dispatches, research
|
|
178
|
+
runs and **active** time across the WHOLE task and is *not* reset by a user message. Both turn
|
|
179
|
+
and task active time exclude `ask_user` waits — billing human idle false-stops a normal steered
|
|
180
|
+
session and teaches people to reset reflexively. The auto-research pipe is exempt from the turn
|
|
181
|
+
budget but charged against the task envelope, so it cannot smuggle 8 helpers per dispatch past
|
|
182
|
+
the outer bound.
|
|
183
|
+
|
|
184
|
+
Either envelope stops before another dispatch/research call and asks one localized Yes/No
|
|
185
|
+
`ask_user` question. A confirmed turn continuation renews its counters/clock in the same tool
|
|
186
|
+
loop. A confirmed task continuation opens one audited tranche, also renews the turn, and
|
|
187
|
+
preserves task tier, assertions, capability packs, label, blockers, and progress. No typed
|
|
188
|
+
`continue` message or slash command is required. `set_task_tier` with `new_task: true` is only
|
|
189
|
+
for genuinely different work and clears task identity/state. This is the guardrail the
|
|
190
|
+
post-mortem run never hit: every steering message reopened the turn window,
|
|
191
|
+
so one workspace stayed open 47 hours on a change that took 13 minutes in a narrow one.
|
|
192
|
+
|
|
193
|
+
On top of the mode sit several qualitative guardrails. **Task triage**: the dispatcher
|
|
194
|
+
classifies the current TASK via the `set_task_tier` tool (`trivial`/`small`/`feature`/`project`)
|
|
195
|
+
and the caps drop to min(mode, tier). The tier is task-scoped and **ratcheted** — it survives
|
|
196
|
+
the user's next message, lowering is free, raising needs a stated `reason` — because a
|
|
197
|
+
turn-scoped tier reset to `feature` on every correction. Three refusals enforce
|
|
198
|
+
proportionality in code rather than prose: a duplicate-dispatch guard (near-identical
|
|
199
|
+
re-dispatches within a turn), a **tier persona gate** (`planner`, `plan-reviewer`,
|
|
200
|
+
`architect`, `security-auditor`, `deep-researcher` are refused at trivial/small — each opens
|
|
201
|
+
a document/finding loop), and a **review round cap** (review dispatches per task, by tier).
|
|
202
|
+
Skipping triage assumes `small`, not `feature`: the tier latches for the task, and the case
|
|
203
|
+
where it was never declared is exactly the case where proportionality was not being
|
|
204
|
+
considered. The review **finding** budget is the one thing deliberately left advisory — the
|
|
205
|
+
hub counts blocking findings (`review-findings.js`) and reports an over-budget return, but
|
|
206
|
+
never reclassifies one, because no rule it can evaluate separates an invented manifest from
|
|
207
|
+
a leaked credential; the round cap is what carries the enforcement.
|
|
208
|
+
**Docs lane** (`docs-lane.js`): a dispatch whose whole declared `scope` is documentation
|
|
209
|
+
refuses review personas (overridable with `review_reason`) and tells the dispatcher not to
|
|
210
|
+
open a review gate; an absent scope is never the lighter lane. **External-blocker stop**
|
|
211
|
+
(`external-blocker.js`): a specialist emits `EXTERNAL_BLOCKED: …` when it needs something
|
|
212
|
+
outside the fleet's reach (account, permission, credential, telemetry destination), and the
|
|
213
|
+
hub refuses the next dispatch with an owner-escalation packet until the human is addressed —
|
|
214
|
+
the alternative, observed, is hours spent approximating the missing fact with scripts,
|
|
215
|
+
manifests and fixtures while the assertion still ends UNPROVEN. **Drift watchdog**
|
|
216
|
+
(`drift-watchdog.js`): armed dispatches are
|
|
217
|
+
observed in-flight from the JSON event stream — deterministic rules (out-of-scope writes
|
|
218
|
+
against the declared `scope` globs, tool-call loops, consecutive failures, tool-call cap)
|
|
219
|
+
escalate to a one-shot cheap LLM judge whose DRIFTING/STUCK verdict terminates the run as
|
|
220
|
+
`drift_stop` (exit 125, partial output preserved); enabled per hub/agent/dispatch
|
|
221
|
+
(`watchdog` key, `/af-watchdog`, `watchdog` param). Two rules about scope: the session's
|
|
222
|
+
own `artifacts/`, `findings/`, and `delegations/` subtrees are implicitly in scope (the
|
|
223
|
+
deliverable protocol *orders* specialists to write there, and the judge is told so), and the
|
|
224
|
+
`scope` rule is non-terminal — it reports a drift advisory on the result and never stops a
|
|
225
|
+
run by itself, matching the post-run scope gate, which reverts nothing. **Dynamic teams**: `/af-agents-add`,
|
|
226
|
+
`/af-agents-drop`, `/af-agents-save` restructure the roster live (the system prompt rebuilds
|
|
227
|
+
every turn), and the gated `team_adjust` tool lets the dispatcher itself adjust the roster
|
|
228
|
+
outside fast mode, with user notification. `/af-hub-report` accounts each turn's dispatches,
|
|
229
|
+
tokens (billed = input + cacheRead + cacheWrite), recycles, drift stops, and refusals.
|
|
230
|
+
|
|
231
|
+
## Runtime stack (tools the fleet sits on)
|
|
232
|
+
|
|
233
|
+
```mermaid
|
|
234
|
+
flowchart TD
|
|
235
|
+
AF["<b>Agent Fleet</b><br/>agent-hub · personas · skills · coms · bridges · CLI"]
|
|
236
|
+
AF -->|primary runtime| PI["<b>pi</b><br/>coding agent — loads harnesses,<br/>extensions, prompts, personas"]
|
|
237
|
+
AF -->|control plane| HERDR["<b>herdr</b><br/>tiled peer workspaces,<br/>presence, snapshot/resume"]
|
|
238
|
+
AF -->|coms peer only| CC["<b>Claude Code</b><br/>bidirectional peer via<br/>the coms bridge<br/><i>not an install target</i>"]
|
|
239
|
+
AF -->|remote human| HERMES["<b>Hermes</b><br/>hub questions relayed to your phone,<br/>plus the Desktop fleet panel"]
|
|
240
|
+
AF -->|outbound remote delegation| CODEX["<b>Codex Remote Control</b><br/>Android-approved calls to<br/>listed coms peers (experimental)"]
|
|
241
|
+
```
|
|
242
|
+
|
|
243
|
+
### External dependencies
|
|
244
|
+
|
|
245
|
+
These are the external systems Agent Fleet assumes or integrates with — not npm packages, but the **runtime stack** the fleet operates on top of.
|
|
246
|
+
|
|
247
|
+
| Dependency | Role | Required? |
|
|
248
|
+
| --- | --- | --- |
|
|
249
|
+
| **[pi](https://github.com/badlogic/pi-mono)** (or your pi install) | Primary coding-agent runtime; loads harnesses, extensions, prompts, and personas | Yes for `just fleet` |
|
|
250
|
+
| **[herdr](https://herdr.dev)** | Workspace control plane: Hub/peer panes, presence push events, team snapshot/resume | Required for `--herdr` or `--peers`; optional for bare `just fleet` |
|
|
251
|
+
| **[Claude Code](https://docs.anthropic.com/en/docs/claude-code)** | First-class coms peer via the [coms bridge](claude-code-coms-bridge.md) — cross-model review and analysis. Never an install target: it hosts no skills, commands, or personas | Optional peer |
|
|
252
|
+
| **Hermes** | Remote human-in-the-loop (Telegram relay for hub questions — [coms-hermes-bridge](coms-hermes-bridge.md)) and the Desktop fleet panel ([hermes-desktop-plugins](hermes-desktop-plugins.md), needs v0.19.0+ and the Desktop app) | Optional |
|
|
253
|
+
| **Codex CLI + ChatGPT Android** | Experimental outbound remote-control conductor on supported `0.144.x`; requires Node `22.6+`, user systemd, interactive pairing, and per-command mobile approvals — [runbook](codex-remote-conductor.md) | Optional / revalidate after minor-version or mobile-client changes |
|
|
254
|
+
| **[addyosmani/agent-skills](https://github.com/addyosmani/agent-skills)** | Upstream skill library (manually vendored) | Bundled (vendored) |
|
|
255
|
+
| **[disler/pi-vs-claude-code](https://github.com/disler/pi-vs-claude-code)** | Source inspiration / MIT port origin for pi harnesses | Design lineage (ported in-repo) |
|
|
256
|
+
| **LLM providers** | Models per persona (`model:` / `models:` in agent frontmatter) — e.g. OpenAI Codex, GitHub Copilot, Ollama, … | Yes (at least one provider your agents can call) |
|
|
257
|
+
| **Chrome DevTools MCP** / **Playwright Agent CLI** | Browser verify (`browser-testing-with-devtools`) and headless automation (`bowser`) | Optional, feature-specific |
|
|
258
|
+
| **Node.js + npm** | CLI (`npx @chankov/agent-fleet`), package install, `just` recipes | Yes for install & tooling |
|
|
259
|
+
|
|
260
|
+
## Repository module map
|
|
261
|
+
|
|
262
|
+
```text
|
|
263
|
+
.pi/ # Pi runtime: harnesses, extensions, agents config, prompts
|
|
264
|
+
skills/ # Agent Fleet-native skills (shadow vendored names)
|
|
265
|
+
agents/ # Personas/subagents used by Agent Fleet
|
|
266
|
+
scripts/ # CLI helpers, bridges, team + one-off peer launchers (pure logic in scripts/lib/)
|
|
267
|
+
hermes/ # Hermes-facing skills/integration assets
|
|
268
|
+
codex/ # Canonical Codex conductor contract (runtime copy lives outside checkout)
|
|
269
|
+
systemd/user/ # Owned user-unit template for Codex remote control
|
|
270
|
+
vendor/agent-skills-upstream/ # Manually imported upstream skills (pinned SHA)
|
|
271
|
+
bin/ # npm CLI: init/update/doctor + the installer engine
|
|
272
|
+
hooks/ # The coms bridge Stop hook
|
|
273
|
+
references/ # Supplementary checklists, installed with the skills that cite them
|
|
274
|
+
docs/ # This file, setup guides, bridge references, vendoring policy
|
|
275
|
+
```
|
|
276
|
+
|
|
277
|
+
Reserved for future modules (do not repurpose these paths):
|
|
278
|
+
|
|
279
|
+
```text
|
|
280
|
+
apps/dashboard/ # future dashboard: Kanban state, Herdr workspaces, peer status
|
|
281
|
+
packages/fleet-core/ # future extracted core orchestration library
|
|
282
|
+
packages/herdr-bridge/ # future Herdr integration package
|
|
283
|
+
packages/hermes-bridge/ # future Hermes integration package
|
|
284
|
+
```
|
|
285
|
+
|
|
286
|
+
## Design rules
|
|
287
|
+
|
|
288
|
+
- **Thin dispatcher context.** Nothing lands persistently in the dispatcher's
|
|
289
|
+
context if it can live on disk or in a one-line status. Research findings,
|
|
290
|
+
the Verification Contract ledger, and team snapshots are all disk-first.
|
|
291
|
+
- **A harness fault must never look like a specialist fault.** The hub passes
|
|
292
|
+
`--session <file>` on every run, so one corrupt session file used to fail a
|
|
293
|
+
persona in ~1s with no output — indistinguishable from a bad agent, and
|
|
294
|
+
unrecoverable by drop + re-add. Unusable session files are now validated and
|
|
295
|
+
quarantined (`session-health.js`) with the reset named in the result. Same rule
|
|
296
|
+
for the return contract: a report the parser cannot read gets one cheap
|
|
297
|
+
read-only extraction pass (`return-extract.js`) before its assertions are
|
|
298
|
+
written off as unproven, and extracted evidence is always labelled as weaker
|
|
299
|
+
than declared evidence. The converse also holds: a run that errored or timed
|
|
300
|
+
out writes to `artifacts/failures/`, never `returns/` — an error stub filed
|
|
301
|
+
as a return reads as a specialist verdict and gets acted on as one.
|
|
302
|
+
- **A pool status field beats reading the screen.** `coms_list` publishes each
|
|
303
|
+
peer's `pane_id` and `status` (`idle`/`working`/`booting`), and
|
|
304
|
+
`herdr_spawn_peer` waits for the peer to register and returns `peer_ready`
|
|
305
|
+
rather than a bare pane id. A spawned peer boots idle and does nothing until
|
|
306
|
+
addressed, so peers spawned and never sent to are named at turn end and in
|
|
307
|
+
`/af-hub-report` (`spawned-peers.js`); closing stays the human's call.
|
|
308
|
+
- **A declared requirement carries its origin.** Every assertion in the ledger
|
|
309
|
+
names its source (`assertion-ledger.js`), and the open ledger is soft-capped
|
|
310
|
+
at 8 — an id nobody can trace back to a plan line costs a dispatch and an
|
|
311
|
+
ASK_USER cycle to re-derive.
|
|
312
|
+
- **Pre-flight validation is free.** Anything the hub can reject before spawning
|
|
313
|
+
— an unresolvable artifact path, an unknown research persona, a heavy persona
|
|
314
|
+
at a low tier, a reviewer on a docs-only scope — is refused without spending a
|
|
315
|
+
budget slot. Artifact paths also resolve across artifact kinds when the name is
|
|
316
|
+
unique, since the hub writes every auto-return under `returns/` while
|
|
317
|
+
dispatchers reasonably guess `reviews/`.
|
|
318
|
+
- **Evidence is archived, never overwritten.** Session start moves the previous
|
|
319
|
+
session's `artifacts/` into an immutable `.pi/agent-sessions/runs/<runId>/`
|
|
320
|
+
namespace with a read-only `meta.json` and an appended `runs/index.json`
|
|
321
|
+
(`run-namespace.js`), retained per `run-history-keep` (default 10). The old
|
|
322
|
+
behaviour — delete `artifacts/` at start, name returns by a per-session counter
|
|
323
|
+
— collided two ways at once and made a post-mortem record eleven specialist
|
|
324
|
+
returns and two reviews as NOT RECOVERABLE. A failed archive leaves the
|
|
325
|
+
artifacts in place: a stale tree is recoverable, a deleted one is not.
|
|
326
|
+
- **Herdr owns panes, presence, and lifecycle; coms owns messages.** Herdr
|
|
327
|
+
topology recipes (`--herdr` or `--peers`) require a running server and refuse
|
|
328
|
+
with an actionable message otherwise. Bare Fleet does not require Herdr.
|
|
329
|
+
`--no-coms` leaves direct/native execution available but disables peer
|
|
330
|
+
messaging, peer-backed dispatch, and handoff from the Hub.
|
|
331
|
+
- **External agents are peers, not plugins.** Claude Code (and future CLI
|
|
332
|
+
agents) join the fleet through bridge adaptors that speak coms envelopes —
|
|
333
|
+
the fleet core stays agent-agnostic. Hermes remains the inbound `ask_user`/
|
|
334
|
+
Telegram route; the experimental Codex conductor is outbound-initiated only,
|
|
335
|
+
approval-gated, and restricted to listed peers through the validated wrapper.
|
|
336
|
+
- **Hermes monitor presentation is outside the fleet core.** Agent Fleet exposes owner-only
|
|
337
|
+
local monitor operations for Hub-owned state; a consumer owns its UI and lifecycle. The
|
|
338
|
+
worktree contains additive monitor/event/invoke code, but that implementation and its local
|
|
339
|
+
tests are not proof of a durable external identity or live delivery contract. `invoke`, where
|
|
340
|
+
available, is Hub-owned and queues dispatcher work rather than exposing tools directly.
|
|
341
|
+
- **Packaged Hermes source is opt-in, never auto-installed.** The npm tarball carries the
|
|
342
|
+
`hub-watchdog` skill (`hermes/skills/`) plus the backend and Desktop monitor plugin source
|
|
343
|
+
(`hermes/plugins/`, `hermes/desktop-plugins/`) as runtime-only source. Shipping that source
|
|
344
|
+
makes it available to an operator; installing it into a Hermes profile is always an explicit
|
|
345
|
+
action through `agent-fleet set-hermes-watchdog` (skill) or the consumer's own flow (plugins).
|
|
346
|
+
Nothing is enabled, launched, or configured by installing the package.
|
|
347
|
+
- **Watchdog delivery is capability-gated and currently unproven.** No checked-in
|
|
348
|
+
Gate O live artifact proves Hermes origin identity, updates, reconnect, or two-chat isolation,
|
|
349
|
+
so its supported posture is journal-only/dormant: no delivery, steering, or surgical use. It
|
|
350
|
+
never manages services, gateways, Herdr, or shell commands. Local runtime evidence — including
|
|
351
|
+
a real foreground watcher against a disposable Hub UDS — is `synthetic-local` and proves none of
|
|
352
|
+
those capabilities; see [the watchdog runbook](hermes-watchdog-supervisor.md).
|
|
353
|
+
- **External conductor contracts are advisory.** Pi damage-control wraps Pi
|
|
354
|
+
tool calls, not Hermes or Codex processes; human approvals and their
|
|
355
|
+
contracts reduce risk but do not provide an OS command allowlist.
|
|
356
|
+
- **Destructive fleet verbs are damage-control-guarded.** Specialists cannot
|
|
357
|
+
spawn/close herdr panes; the human confirms destructive actions.
|
|
358
|
+
- **Native-over-vendored skills.** The skill catalogue resolves `skills/`
|
|
359
|
+
first, then the vendored upstream import; upstream updates are explicit
|
|
360
|
+
maintainer actions ([UPSTREAM-SKILLS.md](UPSTREAM-SKILLS.md)).
|
|
361
|
+
|
|
362
|
+
## History
|
|
363
|
+
|
|
364
|
+
Agent Fleet began as a fork of `addyosmani/agent-skills` and was split into a
|
|
365
|
+
standalone repository in July 2026, with upstream demoted to vendored content.
|
|
366
|
+
The one-time migration record, including the history-filtering commands, lives
|
|
367
|
+
in [MIGRATION-agent-fleet.md](MIGRATION-agent-fleet.md); the product
|
|
368
|
+
requirements that drove the split are in
|
|
369
|
+
[prd-agent-fleet-split.md](prd-agent-fleet-split.md).
|