ultra-task

ultra-task is a skill for Claude Code, Codex from KonghaYao/peri. It costs 75 tokens per session (1,577 once invoked), scanned A, original, Apache-2.0.

A task-supervision method that makes the main coding agent coordinate smaller helper agents. It keeps responsibility for the user's goal, permissions, overall design, and final checks with the main agent.

In plain words
What is it for?
Use it for non-trivial repository work that benefits from parallel investigation or implementation, with the main agent integrating and verifying the results.
Why use it?
It reduces the effort of handling a large task by splitting bounded investigation, coding, testing, and review work among helpers while tracking dependencies and acceptance evidence.

Skill for Claude CodeCodex

Written for no agent in particular: nothing here depends on one. Also seen: mentions subagents.

Good fit Use it for non-trivial repository work that benefits from parallel investigation or implementation, with the main agent integrating and verifying the results.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/konghayao/peri/ultra-task
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add KonghaYao/peri --skill ultra-task
Clone the repo
git clone --depth 1 https://github.com/KonghaYao/peri

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for ultra-task

README.md
[![agentmods](https://agentmods.dev/badge/skills/konghayao/peri/ultra-task.svg)](https://agentmods.dev/skills/konghayao/peri/ultra-task)
Your own site
<a href="https://agentmods.dev/skills/konghayao/peri/ultra-task"><img src="https://agentmods.dev/badge/skills/konghayao/peri/ultra-task.svg" alt="Measured on agentmods" height="20"></a>
Per session 75 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,577 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00075 $0.01577
Opus 5 $0.00037 $0.00788
Sonnet 5 $0.00015 $0.00315
Haiku 4.5 $0.00007 $0.00158

Measured 2d ago against content hash cdd93753bded, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

ultra-task scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

peri-middlewares/src/skills/builtin/skills/ultra-task/SKILL.md · 159 lines

How it starts

The opening of the file, as written. The whole thing — 159 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Ultra Task

Act as the task supervisor. The Main Agent is the control plane: it owns the user's intent, task graph, authorization boundary, integration, and completion verdict. Subagents are the execution plane: delegate most bounded investigation, implementation, testing, and review work to them.

An explicit /ultra-task invocation always selects this mode. Delegation is not a quota: keep decisions that change user intent, permissions, scope, or global architecture in the Main Agent, and perform tiny coordination or integration actions directly when delegating them would cost more than doing them.

Use the direct Agent tool. Do not switch to Workflow or Ultra-ADLC unless the user separately asks for that mode. Subagents cannot launch further subagents.

1. Establish control

Before dispatching work:

  1. Read the repository instructions and inspect enough primary evidence to identify the real task boundary.
  2. Translate the request into acceptance requirements and a dependency-aware ledger of work packages.
  3. Record for each package: objective, dependencies, owner, read/write scope, expected artifact, acceptance evidence, and state.
  4. Map every acceptance requirement to a package or to an explicitly retained Main-Agent responsibility.

Keep the ledger in the current plan unless the user requested a durable artifact. Use these states: pending, running, candidate, accepted, and blocked. A subagent result is only candidate until the Main Agent checks its artifacts and evidence.

Ask the user before fan-out only when a missing product decision or authority would materially change the result. Do not ask them for facts available in the workspace.

This step is complete when all acceptance requirements have an owner, observable evidence, and non-overlapping write ownership.

2. Issue work contracts

Delegate packages that are ready according to their dependencies. Each Agent prompt is a work contract and includes:

Objective: one bounded outcome
Context: only the facts needed to act correctly
Ownership: exact files, modules, or result dimensions this agent owns
Boundaries: what it must not change or decide
Inputs: repository instructions, dependencies, and prior results to consume
Acceptance: commands, observations, or artifacts that prove the work
Budget: search depth, effort or output limit, and an explicit stop condition
Return: concise result, files changed, evidence, blockers, and remaining work
Coordination: other agents may be editing; preserve unrelated work and never revert it

Read the full file on GitHub · 159 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 159 lines · 75 tokens per session scan A cdd93753bded

Subscribe to this mod's changes

ultra-task is a skill published in the GitHub repository KonghaYao/peri (160 stars, last pushed 2d ago), licensed Apache-2.0. It adds 75 tokens to every session and 1,577 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-05.

Related

Other skills, from other repositories

qa-testing

Verify your work by actually operating the app or website you changed, instead of assuming it works. Strongly recommended whenever you build, modify, or debug a web app, website, or desktop GUI app. Drive real browsers with the agent-browser CLI and native desktop apps with the cua-driver CLI. These are installed on…

openinterpreter/openinterpreter · 77 tokens

plugin-creator

Create and scaffold plugin directories for Codex with a required .codex-plugin/plugin.json, optional plugin folders/files, valid manifest defaults, and personal-marketplace entries by default. Use when Codex needs to create a new personal plugin, add optional plugin structure, generate or update marketplace entries…

openinterpreter/openinterpreter · 86 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openinterpreter/openinterpreter · 113 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openinterpreter/openinterpreter · 114 tokens

skill-creator

Create or update a Codex skill with appropriately scoped instructions and any needed supporting resources.

openinterpreter/openinterpreter · 21 tokens

codex-pr-body

Update the title and body of one or more pull requests.

openinterpreter/openinterpreter · 17 tokens