5efe97c0ae
Extends the doc from #3255/#3273 with the failure that cost the most in a single session: **one stale install produced five wrong reports on one issue** (#3264). ## What happened A `node_modules` that had drifted from the lockfile — `jsdom@29.0.1` installed, `29.1.1` pinned — generated failures that existed on no CI machine and no other checkout. They were not subtle: deterministic, reproducible on demand, with plausible stack traces and real-looking assertion diffs. Each round of triage got **more precise about the wrong data**: | round | claim | why it was wrong | | --- | --- | --- | | 1 | "4 deterministic failures" | measured in a 4-file batch, called it isolation | | 2 | "3 deterministic, 2 order-dependent" | isolated correctly, but a race is not deterministic | | 3 | "TaskCard is broken" | stale jsdom; the CSS assertion was correct | | 4 | "no contamination" | true of four app files; published unqualified | | 5 | "quarantine these two" | never read the failure text — both were timeouts | The through-line is not carelessness about the code. **The environment was never treated as part of the claim**, so no amount of care about the analysis could recover it. ## The checks, in the order they cost the most ```bash pnpm install --frozen-lockfile # node_modules is not evidence until it matches the lockfile <run the file ALONE, 3+ times> # isolation and repetition answer different questions <read the failure TEXT> # a timeout and an assertion failure need opposite responses uptime # a loaded box manufactures timeouts that mean nothing ``` ## Why the load check earned its place Two tests "failing" in a full-suite run were `Test timed out in 15000ms` on a box at **load average 9.7 with 84 users**. Under AGENTS.md's quarantine-on-sight rule that reads as a flake to quarantine — and the ledger's **14-day deletion ratchet would have made the lost coverage permanent**. The rule presumes the failure is a property of the test, not of the machine. A wall-clock budget crossed under local contention is evidence about the hardware. I was one comment away from deleting healthy coverage on that basis. ## The tell A finding is environment-derived when it is **local, recent, and unshared**: nobody else has reported it, CI is green, and it appeared without a commit that could explain it. Any two of those should stop a report before it is written. All three applied here, and the report went out anyway — five times. ## Verification Docs only; no code paths change. `fnxc-future-dates`, `lifecycle-columns`, `quarantine-ledger` exit 0. No changeset — internal docs are excluded. Context: the one finding in #3264 that survived all five rounds is #3286 (merged), and it survived because it was verified by **reverting the product change** rather than by trusting a red — 3/3/2 failures without the fix, 27/27 across four runs with it.
Fusion Documentation
Fusion is an AI-orchestrated task board that turns ideas into reviewed, merged code using a structured workflow: planning → todo → in-progress → in-review → done.
Quick Start
Start the local dashboard with pnpm dev dashboard, then create your first task from the board or CLI.
For a full walkthrough (installation, onboarding, first task, and daily workflow basics):
Documentation Index
Getting Started
| Guide | Description |
|---|---|
| Getting Started | Installation, first-run, first task, and daily workflow basics |
| Dashboard Guide | Board/list views, left/right sidebar navigation, Artifacts, Import Tasks, chat, workflow selection/editor, terminal, git manager, files, planning, and UI tools |
| CLI Reference | Complete fn command reference with subcommands, flags, and examples |
| Remote Access | Operator runbook for Tailscale/Cloudflare setup, tokenized login links, security caveats, and troubleshooting |
| Native Shell Connection Guide | Canonical mobile/desktop shell onboarding, profile management, QR/manual setup, and remote handoff behavior |
Task & Project Management
| Guide | Description |
|---|---|
| Task Management | Task creation modes, lifecycle, prompt specs, comments, archiving, and GitHub integration |
| Todo View | Canonical guide for the experimental Todo View, including enablement, usage, API routes, and storage |
| Missions | Mission hierarchy, planning flow, activation, progress tracking, and autopilot behavior |
| Goals Refinement Gate | Evidence gate for activating the conditional post-v1 goals refinement slice only after real usage pain is documented |
| Goals Refinement Evidence Pack | Structured observation template and two-observation threshold for conditional Slice 4 activation requests |
| Research | Research runs, provider setup, dashboard/CLI usage, findings, exports, and task integration |
| Research View UX Spec | Canonical layout and capability-state messaging spec for the Research dashboard view (FN-4138, informs FN-4134/FN-4135) |
| Workflow Steps | Workflow overview, built-in workflow catalog, per-task selection, runtime semantics, reusable quality gates, templates, phases, and execution results |
| Workflow Editor | Visual workflow editor guide for opening, viewing, authoring, validating, importing/exporting, custom fields/columns/settings, and tuning workflows |
| Custom Workflow Reliability Acceptance Map | End-to-end reliability acceptance criteria for custom workflow authoring, selection, execution, recovery, restart durability, and deferred journeys |
| Custom Non-Coding Workflows MVP Spec | MVP framing for user-authored non-coding workflows, lifecycle mapping, metrics, and risk checklist |
| Task Evaluations | Eval scoring contract, evidence persistence, score categories, and evaluation pipeline |
| Multi-Project | Central registry architecture, project management, isolation modes, and migration paths |
Configuration & Agents
| Settings Reference | Global/project settings, workflow setting values, model/fallback lane hierarchy, defaults, and API endpoints |
| MCP | Model Context Protocol server configuration, secret references, validation, CLI, dashboard, and import/export workflows |
| Agents | Agent management, presets, prompts, heartbeat behavior, spawning, and mailbox workflows |
| Planner Oversight (see Settings Reference, Dashboard Guide, Architecture) | Workflow-native oversight levels (off/observe/steer/autonomous), per-task overrides, notification verbosity, the human-confirmation gate on merge/PR and destructive actions, and the Task Detail overseer controls/Intervention Timeline |
Architecture & Development
| Guide | Description |
|---|---|
| Architecture | System architecture, package layout, storage model, and engine execution flow |
Secrets Store (SecretsStore) |
Core encrypted secret subsystem overview: scopes, AES-256-GCM at-rest model, policy semantics, and public store API surface |
| Dashboard Real-Time | Canonical event-stream architecture contract (shared /api/events bus + dedicated stream boundaries), with project/node scoping, reconnect/cleanup behavior, and realtime pitfalls |
| Storage | PostgreSQL runtime storage, archive, migration compatibility, and file-backed payloads |
| DAG Architecture Deliverables | Milestone A DAG architecture documents plus Milestone B prototype scaffold docs (schema migration plan, DagCoordinator design, implementation checklist) |
| Dev Server Module Audit | Analysis of parallel dashboard dev-server module families, production wiring, and consolidation guidance |
| Shared Cluster Protocol | Shared PostgreSQL multi-node contract: claims/leases, membership, auth, and retired multi-leader mesh replication |
| Signals Connectors | HMAC-signed external signal connectors for setup, payload mapping, and security notes across Sentry, Datadog, PagerDuty, and generic webhooks |
| Multi-Project Sequencing and Dependency Analysis | Sequencing guidance for FN-3448/FN-3449/FN-3503/FN-3182, including identity boundaries and recommended board dependency edges |
| Contributing | Local development setup, testing, release flow, and contributor conventions |
| Docker | Container builds, deployment, and persistence configuration |
| Code Signing | macOS and Windows code signing configuration for release binaries |
| Diagnostics | Engine diagnostic logging subsystems, structured log keys, and key diagnostic points catalog |
| Sandbox Backends | Pluggable sandbox backends for executor command isolation (bubblewrap, spawn-based) |
| Secrets | Encrypted secrets storage, per-secret access policies, scopes, and agent tool wiring |
| Testing | Full testing lanes, worker fanout guidance, test taxonomy, and file organization |
| Real iOS Safari Acceptance Surface | Provisioning runbook and harness usage for terminal verification gates on physical or cloud real-iOS Safari |
| Solutions Catalog | Documented solutions to past problems (bugs, architecture patterns, best practices) organized by category |
| Localization Contributing Guide | Conventions for contributing translations, locale file structure, and i18n tooling |
| Mobile | Capacitor/PWA mobile development setup and workflow |
Plugins
| Guide | Description |
|---|---|
| Plugin Management | End-user guide for discovering, installing, enabling, configuring, updating, uninstalling, and troubleshooting Fusion plugins |
| Plugin Authoring | Developer guide for building Fusion plugins (manifest, SDK hooks, routes, UI/runtime contributions) |
| Even Realities Glasses Plugin | Task-focused Even Realities glasses bridge with quick capture, polling notifications, and agent actions |
| Reports Plugin | Reports plugin rendering, export, standalone HTML generation, and section configuration |
| Even Realities Plugin API | Even Realities plugin API endpoint reference and test coverage matrix |
| Memory Plugin Contract | Pluggable memory backend architecture, interface contract, and migration strategy |
| Compound Engineering Plugin | CE workflow dashboard surface: artifact hub, interactive sessions, work→board bridge, and bidirectional sync |
| External Plugin Authoring | Step-by-step guide for authoring plugins using an installed fn CLI (no monorepo access needed) |
| External Plugin Proof-Point Runbook | Repeatable release-validation runbook for proving an external plugin runs against a published Fusion CLI build |
Audit Reports
| Report | Description |
|---|---|
| Test Feedback-Loop Baseline | Weekly FN-6612 signal-per-second baseline for gate/test wall-time, slowest files, and quarantine trends |
| Test Value Audit | Heuristic test-value audit generated by scripts/test-value-audit.mjs to support human deletion and review decisions |
| Test Velocity Baseline | Weekly feedback-loop velocity baseline for merge-gate, boot-smoke, changed-test, and quarantine metrics |
| UX Audit Report | Comprehensive UX audit with prioritized recommendations for dashboard improvements |
| Codebase Improvement Audit | Evidence-based technical debt and reliability gap audit with prioritized recommendations |
| Gap Analysis | System completeness analysis comparing Fusion to Paperclip feature set |
| Permanent Agent Heartbeat Playbooks | Worked manager/IC/message/blocked/no-task heartbeat scenarios and anti-patterns |
| Agent Sandbox Research | Research on agent isolation, capability enforcement, and sandboxing approaches |
| Even Realities Integration Research (FN-3737) | Research summary and recommended integration topology for Even Realities glasses + Fusion |
| pi-autoresearch Analysis for Fusion Port | Upstream architecture/license analysis and Fusion integration mapping for autoresearch capabilities |
| pi-autoresearch Audit vs Fusion Research | Audit comparing Fusion's research subsystem against upstream pi-autoresearch capabilities and parity gaps (FN-4136) |
| Research Hardening Preflight Baseline | Verified research subsystem baseline, lifecycle contracts, and hardening pressure points |
| Test Audit Report | Test coverage and effectiveness audit with recommendations |
| Skipped Test Inventory | Current intentional test-skip inventory and reconciliation status for older skip follow-ups |
| Dev Server Module Boundary Audit | Boundary/ownership audit for parallel dev-server-* vs devserver-* dashboard modules and FN-2212 prioritization guidance |
| spawn_agent Approval Evaluation (FN-3973) | Decision to keep fn_spawn_agent under generic action-gate governance rather than durable agent provisioning policy |
| Task Lineage Reconciliation Notes | Historical task-ID reuse patterns, confidence semantics for commit attribution, and reconciliation methodology (FN-3953, FN-3998) |
| Dashboard Load Performance (historical) | Pre-cutover SQLite index analysis retained for performance archaeology |
| CLI Printing Press Plugin Design | Architecture design for the CLI printing press bundled plugin (FN-3762) |
| CLI Printing Press Research | Upstream cli-printing-press analysis and Fusion integration mapping (FN-3761) |
| Research vs Experiment Session Naming Decision | Naming decision record: hybrid approach retaining research_* for cited-search/synthesis and adding experiment_session_* for upstream parity (FN-4223) |
| Experiment Executor Design | Experiment executor architecture: lifecycle, run state machine, and worktree isolation model |
| Experiment Finalize Flow | Experiment finalize contract: branch grouping, dry-run planning, and session completion semantics |
| Experiment Session Model | Experiment session data model: state transitions, iteration tracking, and persisted run state |
| Experiment Session MVP Spec | MVP specification for the experiment session feature: scope, invariants, and delivery milestones |
| Sandbox Options Research (FN-4635) | Pluggable sandbox options research: threat model, backend evaluation, and spawn-based isolation design |
| Triage Duplicate Detection Postmortem | Postmortem on duplicate task detection gaps and scheduler dedup hardening |
| Multi-Node Runtime Readiness (FN-4814) | Runtime readiness assessment for multi-node distributed coordination |
| Distributed Multi-Node Coordination Gap (FN-4819) | Gap analysis for distributed multi-node agent coordination and cross-node task assignment |
| Cross-Node Assignment Wake Contract (FN-4824) | Contract specification for cross-node task assignment wake signaling |
| Multi-Node Coordination Validation Findings (FN-4820) | Validation findings from multi-node coordination testing and edge-case analysis |
| Secrets Sync Auth Parity Review (FN-4886) | Review of node secrets sync API authentication parity and security boundaries |
| Test Speed Audit (FN-5048) | Measured baseline test performance, offender list, and optimization priorities |
| Soft-Delete Verification Matrix | Authoritative checklist for the FN-5105 → FN-5143 soft-delete stream: scenario × layer coverage |
| Self-Healing Backward Move Audit | Audit of self-healing backward-move safety checks and edge-case validation |
| Workflow Policy Ownership Map | U1 characterization map classifying production merge, retry, scheduling, and recovery policy branches before workflow-policy migration cutover |
| Test-Speed Baseline (2026-06-03) | Measured per-file test timing baseline and optimization targets (successor to FN-5048 audit) |
| ACP Runtime Contract | Agent Client Protocol plugin launch/readiness contract and failure taxonomy |
| ACP MCP Passthrough & Permission Forwarding Upstream Sponsorship (FN-6475) | Ready-to-file upstream sponsorship for claude-code-cli-acp ACP session/new.mcpServers passthrough and permission-gate traversal; Route A remains NOT GO until proven |
| Mission Completion Gate Contract | Decision record for mission completion gate invariants and acceptance flow |
| Lost-Work Tasks Incident (2026-05-23) | Incident catalog of 9 lost-work tasks from no-op finalize and reuse-handoff bugs | | GitLab Parity Inventory (FN-7421) | Implementation map for first-class GitLab support: import, linked issue tracking, comments, auth/settings UI, CLI/extension, and Command Center surfaces to mirror or explicitly exclude | | PostgreSQL Runtime Cutover Review (2026-07-14) | Current end-to-end authority inventory, intentional legacy SQLite readers, deployment contract, and verification record | | SQLite → PostgreSQL Migration Review (2026-06-26, historical) | Historical multi-agent review of the incomplete migration branch and its original findings | | Dashboard Theme & UI Plugin System Proposal (2026-07-01) | Feasibility-spike proposal for a controlled dashboard theme/UI shell extension point sharing one backend source of truth | | Full-loop Agent Tool-Surface Audit and Delivery Plan | Source-grounded audit of engine-agent and dashboard chat tool factories, gap analysis for mission hierarchy integration, and delivery plan (FN-8280) | | Dashboard Modal Inventory | Canonical classification of all 45 dashboard modal surfaces (classes A–D) with file:line evidence, FloatingWindow migration targets, and the shared migration contract (FN-8605 → FN-8617) | | Workflow-Owned Lifecycle Closing Verification | Closing-bar verification runbook and recorded pass history for the workflow-owned lifecycle cutover programme — gate, verify:fast, E2E families, and census |
External Resources
- GitHub repository: https://github.com/Runfusion/Fusion
- npm package: https://www.npmjs.com/package/@runfusion/fusion
- pi agent framework: https://github.com/earendil-works/pi
Suggested Reading Paths
- New user: Getting Started → Dashboard Guide → Task Management
- Workflow author: Dashboard Guide → Workflow Editor → Workflow Steps → Settings Reference
- Power user / automation owner: Settings Reference → Workflow Steps → Agents → Planner Oversight (Settings Reference § Workflow Settings)
- Maintainer / contributor: Architecture → Multi-Project → Contributing