skills.rest
volter-ai
volter-aiOFFICIAL
Explore all skills by this author →
GitHub

manager

Executes ready roadmap tasks by dispatching implementation waves, reviewing changes, and landing merged PRs.

Updated Sep 2, 2026
Search Tags:#code review#pull request#github#workflow orchestration#autonomous agents#task execution#ci gating
One-click install
npx skills add https://github.com/volter-ai/open-autonomy-compiler --skill manager-volter-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: manager
Source: https://github.com/volter-ai/open-autonomy-compiler/tree/main/profiles/simple-gh/skills/manager
Command: npx skills add https://github.com/volter-ai/open-autonomy-compiler --skill manager-volter-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating autonomous code changes safely is hard: someone must pick up approved tasks, dispatch implementation agents, enforce review gates, and land changes without ever letting one agent both write and approve code. This Skill operationalizes that execution loop under deterministic guardrails. ## Core Features & Use Cases - Task consumption via task service: Reads ready-state tasks through the configured tracker tool, validates acceptance criteria, dependencies, and risk boundaries before dispatch. - Bounded implementation waves: Dispatches one isolated-worktree implementation subagent per wave, armed with a ztrack loop that enforces the review-state transition. - Review-and-land gating: Requires green checks plus a fresh read-only review pass on the exact head SHA before merging, with at most two rework waves before escalating to a maintainer. - Use Case: A maintainer approves a roadmap task in the tracker; on the next tick the manager dispatches an implementation agent, waits for CI, runs an independent review, merges the PR, and marks the task done with recorded evidence. ## Quick Start Ask the agent to run one manager tick that picks the highest-priority ready task from the configured tracker and drives it through implementation, review, and merge.

Frequently Asked Questions about manager

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How does the manager skill execute ready tasks autonomously?▼

It queries the configured task tool for tasks in the mapped ready state, selects the highest-priority one, dispatches an implementation subagent in an isolated worktree, then reviews and merges the resulting PR once checks and review pass.

How do I configure task states and models for the manager?▼

Configuration lives in .open-autonomy/autonomy.yml: policy.tracker.tool selects the task service, policy.taskStates maps portable lifecycle names, and policy.models sets implementation and research tiers. Dispatches must use the declared tier or fail closed.

Can the manager create new roadmap tasks or reprioritize work?▼

No. The manager only executes tasks already in the ready state; Planner creates and prioritizes product tasks and Kaizen handles process improvements. The manager consumes the task API and never reads persistence files or plan documents.

What happens when a task touches human-required paths?▼

The manager moves the task to the mapped inputRequired state, engages the declared Maintainer through the Runner with the exact decision needed, and stops. It never weakens the risk gate to make work land.

Why does a review become invalid after a new push?▼

Reviews are recorded against a specific head SHA, so any later push invalidates the earlier pass. The manager waits for required checks on the new head and dispatches a fresh read-only review before merging.

How many rework attempts does the manager allow before escalating?▼

At most two rework waves per task. After that, the manager moves the task to the mapped inputRequired state with the accumulated evidence and the maintainer decision needed.

Related Skills

Explore More Software Engineering Skills→
Samuelca6399Samuelca6399

product-strategy

Draft product vision, outcome-based roadmaps, and prioritize initiatives with RICE, ICE, and MoSCoW.

Community
Advanced
tilmon-engineeringtilmon-engineering
3

comparing-branches

Compare two autonomy branches and report divergence, iterations, metrics, and outcomes.

Official
Intermediate
HughYauHughYau
2.5k

self-awareness

Query Claude Science session metadata, token usage, and execution history via SQLite.

Community
Advanced
liumeixinliumeixin
4

github-pr-workflow

Automate the GitHub Pull Request lifecycle using gh CLI or git and REST API.

Community
Advanced