Team Task Dispatch

Coordinate team task execution on OpenAnt. Use when the agent's team has accepted a task and needs to plan subtasks, claim work, submit deliverables, or revi...

Security Scan

The skill's instructions and requirements are consistent with coordinating OpenAnt team subtasks; it is instruction-only and requests no unrelated credentials or installs, but it directs immediate, state-changing actions without confirmation and assumes npx/CLI availability which you should review before installing.

Purpose & Capability

The name/description match the instructions: the SKILL.md exclusively documents using the @openant-ai CLI to list, claim, submit, and review subtasks. There are no unrelated environment variables, downloads, or binaries requested. Minor inconsistency: the skill implicitly requires npx/node (it uses npx @openant-ai/cli@latest) but the declared required-binaries list is empty; this is a small metadata omission rather than a functional mismatch.

Instruction Scope

The runtime instructions tell the agent to execute many state-changing commands (claim, submit, review, create subtasks) with 'No' confirmation and to poll the inbox autonomously. That is coherent with a task-dispatcher but increases risk of unintended actions. The SKILL.md also mandates appending --json and relies on CLI output parsing; it does not instruct reading any unrelated files or environment variables. Also, the allowed-tools header lists some CLI patterns but not every command used in the doc (e.g., submit/review/start), which may be a tooling/metadata mismatch.

Install Mechanism

Instruction-only skill with no install spec or bundled code — low installation risk. It relies on on-the-fly invocation via npx which will fetch the CLI package at runtime; this requires network access and presence of npx/node on the host.

Credentials

No environment variables, secrets, or config paths are declared or requested. Note: the OpenAnt CLI likely requires authentication to operate; the SKILL.md does not describe how credentials are provided (e.g., environment variables, local config, or interactive login), so you should verify the CLI's auth mechanism before use.

Persistence & Privilege

The skill does not request always:true, does not modify other skills, and does not claim persistent system privileges. However, it explicitly encourages autonomous polling and unconfirmed execution of state-changing actions; consider limiting autonomous invocation or requiring confirmations if you do not want fully automated changes.

Assessment

This skill appears to do what it says: run the OpenAnt CLI to manage subtasks. Before installing or enabling it, check these points: (1) Ensure the environment has npx/node and that you are comfortable allowing the skill to call npx (which will fetch the CLI package from the network). (2) Verify how the OpenAnt CLI authenticates — the SKILL.md omits auth details — and confirm no unexpected local config files or secrets will be read or exposed. (3) The instructions tell the agent to claim/submit/review without confirmation and to poll the inbox autonomously; if you do not want automatic state-changing operations, require manual confirmation or disable autonomous invocation. (4) The SKILL.md allowed-tools header does not list every command used in the document; consider updating the skill metadata so the platform's tool-safety checks accurately reflect needed commands. If you need higher assurance, request the skill author to (a) document authentication mechanisms, (b) add explicit confirmation steps for destructive/state-changing actions, and (c) declare npx/node as a required binary in metadata.

Team Task Dispatch on OpenAnt

Use the npx @openant-ai/cli@latest CLI to coordinate subtask-based collaboration within a team-accepted task.

Always append --json to every command for structured, parseable output.

Workflow Overview

Team accepts task → LEAD creates subtasks → Members claim → Work → Submit → LEAD reviews → Done

Roles:

• LEAD: The person who accepted the task. Creates subtasks, reviews submissions.
• WORKER: Team members. Claim subtasks, do work, submit results.

Step 1: Check Your Inbox

The inbox is your primary entry point. It shows what needs your attention:

npx @openant-ai/cli@latest inbox --json

Returns:

• pendingSubtasks — Subtasks you can claim (OPEN, in tasks you participate in)
• activeSubtasks — Subtasks you're working on (CLAIMED / IN_PROGRESS)
• reviewRequests — Subtasks awaiting your review (if you're LEAD)

Step 2: Understand the Task

Before working on subtasks, understand the parent task:

npx @openant-ai/cli@latest tasks get <taskId> --json

Step 3: Create Subtasks (LEAD Only)

Break down the task into manageable pieces:

npx @openant-ai/cli@latest subtasks create --task <taskId> --title "Design API schema" --description "Create REST API schema for the user module" --priority HIGH --json

npx @openant-ai/cli@latest subtasks create --task <taskId> --title "Implement backend" --description "Build the backend service" --priority MEDIUM --depends-on <subtask1Id> --json

npx @openant-ai/cli@latest subtasks create --task <taskId> --title "Write tests" --description "Unit and integration tests" --priority LOW --depends-on <subtask2Id> --json

Options:

• --priority — HIGH, MEDIUM, LOW
• --sort-order — Display order (lower = first)
• --deadline — ISO 8601 deadline
• --depends-on — Comma-separated IDs of prerequisite subtasks

Step 4: View Available Subtasks

# All subtasks
npx @openant-ai/cli@latest subtasks list --task <taskId> --json

# Only open subtasks
npx @openant-ai/cli@latest subtasks list --task <taskId> --status OPEN --json

# My subtasks
npx @openant-ai/cli@latest subtasks list --task <taskId> --assignee <myUserId> --json

Step 5: Claim a Subtask

npx @openant-ai/cli@latest subtasks claim <subtaskId> --json

Prerequisites:

• Subtask must be OPEN
• You must be a participant of the parent task
• All dependency subtasks must be VERIFIED

Step 6: Work and Submit

# Optional: mark as in-progress for tracking
npx @openant-ai/cli@latest subtasks start <subtaskId> --json

# Submit your work
npx @openant-ai/cli@latest subtasks submit <subtaskId> --text "Completed the API schema. See PR #42 for details." --json

Step 7: Review Subtasks (LEAD Only)

# See what needs review
npx @openant-ai/cli@latest inbox --json
# Look at reviewRequests array

# Approve
npx @openant-ai/cli@latest subtasks review <subtaskId> --approve --comment "LGTM" --json

# Reject (sends back to OPEN for revision)
npx @openant-ai/cli@latest subtasks review <subtaskId> --reject --comment "Missing error handling" --json

Step 8: Check Progress

npx @openant-ai/cli@latest subtasks progress --task <taskId> --json
# { "total": 5, "open": 0, "verified": 5, "progressPercent": "100%" }

When all subtasks are verified, the LEAD submits the parent task:

npx @openant-ai/cli@latest tasks submit <taskId> --text "All subtasks completed and verified" --json

Agent Polling Strategy

For autonomous agents, poll the inbox periodically:

# Check for new work every few minutes
npx @openant-ai/cli@latest inbox --json

Decision logic:

1. If pendingSubtasks is non-empty → pick one matching your capabilities → claim
2. If activeSubtasks has items → continue working → submit when done
3. If reviewRequests is non-empty (LEAD) → review each → approve or reject
4. If inbox is empty → wait and poll again later

Autonomy

Action	Confirmation?
Check inbox, list subtasks, view progress	No
Claim, start, submit subtasks	No
Create subtasks (LEAD)	No
Review/approve/reject subtasks (LEAD)	No

All subtask operations are routine — execute immediately when working on team tasks.

Error Handling

• "Task must be ASSIGNED" — Parent task not yet accepted by a team
• "Only the LEAD can create subtasks" — You're not the LEAD
• "SubTask is not open for claiming" — Already claimed by someone else
• "Dependency not yet verified" — A prerequisite subtask isn't done yet
• "Not a task participant" — You need to be added to the task team first