Skip to content

FEATURE SPEC

Questions or comments?

Post them as a comment on the tracking issue -- requires a GitHub account with access to this repo.

Agent-Assisted Task Generation & Distribution

Novaworks AI | Super Agent orchestration (nova-superagent-agent) + x_novaw_platform

Parent Feature Set: ../concept.md.

Major Change History

Version Date Author / Notes
1 2026-08-10 Eswar Vandanapu — Initial draft, derived from row 3 of the parent concept.md's Feature index.

Problem Statement

Even once a Task Force finalizes a charter (Feature 2), nothing bridges that agreed plan into actual work. Today someone still has to manually break the charter into tasks, decide who does what, and chase status by asking around or maintaining a spreadsheet — the exact "reverts to spreadsheets" failure the parent concept.md's business case names, except now it happens after the group already agreed on a plan.

Use Cases

Use Case 1: A Project Owner has just finalized a charter for an upskilling initiative. Today, even with an agreed plan in hand, they'd still have to manually decide what the concrete work items are and email each Task Force member their piece — which matters because the group already did the hard part (agreeing on the plan) and shouldn't lose that momentum translating it into action by hand.

Use Case 2: A Task Force member has been assigned a task from a generated breakdown. Today there's no single place for them to see it or mark it done — status lives in whatever spreadsheet someone's maintaining, if anyone is, which matters because that's precisely the tracking gap this Feature Set exists to close.

Use Case 3: A Project Owner reviews a freshly generated task list and wants to reassign one task to a different Task Force member before anyone is notified. Today no such review checkpoint exists because no tasks are being generated at all — which matters because handing out tasks with zero human check first risks distributing something that doesn't reflect who's actually best suited for it.

Scope

  • Once a Project's charter is finalized (Feature 2), Super Agent reads it and generates a set of concrete execution tasks, each with a suggested assignee drawn from the Project's Task Force (Feature 1) — using the same task-management capability (manage_tasks) nova-superagent-agent already uses for single-user transactions, extended here to multi-person, initiative-level tasks.
  • The generated task list is proposed to the Project Owner for a lightweight review pass — reassigning any task to a different Task Force member — before anything is actually distributed. This is a single batch review, not blind auto-distribution and not a task-by-task approval gate (see Open Questions on whether that granularity needs to widen).
  • Approving the reviewed list distributes each task to its assigned Task Force member, who is notified.
  • Task Force members can update the status of tasks assigned to them (e.g. in progress, done, blocked); the Project Owner can see aggregated status across all of a Project's tasks without asking anyone individually — this is what actually replaces spreadsheet-based tracking.
  • Autonomous: generating the task breakdown and suggested assignees from the charter. Human-governed: the Project Owner's review/reassign pass before distribution, and each Task Force member's own status updates.

Out of Scope

Value

This is the Feature that actually turns an agreed plan into tracked work, closing the planning-to-execution gap the parent concept.md's differentiation section is built around — without it, Features 1 and 2 produce a well-governed plan that still has to be executed by hand.

Delivery

UX Description

A generated task list surfaced to the Project Owner for review within the Project, and individual tasks appearing to each Task Force member in the same task-management surface Super Agent already uses elsewhere (chat, notifications, and a portal task list) — no separate task system, no separate app.

Key User Flows

  1. Feature 2's charter finalizes → Super Agent generates a task breakdown with a suggested assignee per task → presents the proposed list to the Project Owner.
  2. Project Owner reviews the list, reassigns any task to a different Task Force member, approves the batch → each task distributes to its assignee with a notification.
  3. Task Force member opens an assigned task and updates its status (in progress, done, blocked) → the Project Owner sees the Project's aggregated task status update without asking anyone directly.
  4. Super Agent can't produce a confident breakdown from a thin or ambiguous charter → it flags that to the Project Owner with what it does have, rather than distributing a low-confidence task list silently.

Human Override & Fallback Strategy

  • The Project Owner reviews and can reassign every generated task before distribution — no task reaches a Task Force member without having passed through this checkpoint.
  • If the charter is too thin or ambiguous to generate a confident task breakdown, Super Agent surfaces that explicitly to the Project Owner (naming what's unclear) instead of generating and distributing low-quality tasks anyway.
  • Task Force members self-report their own task status; nothing here auto-marks a task done without that person's action.

AI Feedback Loop

  • Explicit: the Project Owner reassigning a suggested assignee before distribution; the Project Owner rejecting/requesting a re-generation of the whole proposed task list if it doesn't reflect the charter well.
  • Implicit: how often generated assignees get reassigned during review (a proxy for suggestion quality); completion rate of generated tasks (a proxy for whether the breakdown was actually actionable, not just plausible-sounding).

Acceptance Criteria

  • Once a Project's charter is finalized, Super Agent generates a set of concrete execution tasks, each with a suggested assignee drawn from the Project's Task Force.
  • The Project Owner can review the generated task list and reassign any task to a different Task Force member before it's distributed.
  • Approving the reviewed task list distributes each task to its assigned Task Force member and notifies them.
  • A Task Force member can update the status of a task assigned to them, and the Project Owner can see aggregated status across all of a Project's tasks without asking anyone individually.
  • Task generation is blocked until the Project's charter is finalized (Feature 2) — this Feature never generates tasks from a draft, unfinalized charter.
  • If Super Agent can't generate a confident task breakdown from the charter, it flags that to the Project Owner instead of distributing a low-confidence task list silently.
  • collaborative-charter-goal-planning (Feature 2): supplies the finalized charter this Feature generates from; task generation blocks until that charter is finalized.
  • project-creation-task-force-assembly (Feature 1): supplies the Task Force members tasks get assigned to.
  • agent-assisted-plan-review-refinement (Feature 5, not yet drafted): a separate, on-demand review/suggestion layer — not part of this Feature's generation/distribution flow.
  • project-reconciliation-replanning (Feature 6, not yet drafted): owns regenerating or adjusting tasks after the fact when the plan changes mid-execution; this Feature only handles the initial generation and distribution.
  • nova-superagent-agent's existing task-management orchestration (manage_tasks): this Feature extends that capability to multi-person, initiative-level tasks rather than building a new task system.

Risks

Risk Severity Mitigation
A generated task breakdown from a thin or ambiguous charter could produce low-quality, actionable-in-name-only tasks Medium The Project Owner review/reassign pass (already in scope) plus explicit low-confidence flagging rather than silent distribution.
Extending manage_tasks to multi-person, initiative-level tasks may need data-model changes (task ownership spanning many assignees on one Project) not yet confirmed feasible on the current implementation Medium Confirm with the team owning nova-superagent-agent before Construction whether multi-assignee, Project-scoped tasks are a natural extension or need new schema.

Open Questions

  • Is the Project Owner's review pass a single batch approval (reassign, then approve all at once), or does it need per-task approval granularity (e.g. editing task descriptions, not just assignees)? Scoped here as a lightweight batch pass; narrower or wider granularity is unresolved.
  • What happens to a distributed task list once the charter changes mid-execution (Feature 6's job) — does this Feature need to support marking previously distributed tasks as stale, or is that entirely Feature 6's responsibility?
  • Figma / Claude Design artifact for the task review/distribution surface is not yet available.

For internal engineering review only.

Novaworks AI | Confidential