fix-task

Orchestrates a four-node investigate, implement, verify, and land workflow for fixing tasks.

14|Updated Mar 2, 2026
One-click install
npx skills add https://github.com/intendant-dev/Intendant --skill fix-task-intendant-dev
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fix-task
Source: https://github.com/intendant-dev/Intendant/tree/main/automations/fix-task
Command: npx skills add https://github.com/intendant-dev/Intendant --skill fix-task-intendant-dev

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating a bug fix across investigation, implementation, verification, and landing often requires manual handoffs and tracking. This Skill automates that pipeline as a dependency graph of scheduled agent sessions, where each node fires only when its prerequisites complete. ## Core Features & Use Cases - Dependency-Driven Nodes: Four nodes (investigate, implement, verify, land) linked by relies_on prerequisites, so each stage starts only after the previous one completes. - Failure Suspension and Re-arming: A repeatedly failing node suspends its own lane until re-approved, and revoking a node halts that lane while downstream nodes stay blocked. - Independent Verification: The verify node re-runs the test battery and live checks independently, completing only on proof. - Use Case: A team describes a bug in the workflow hub; an agent investigates the root cause, another implements the fix and runs tests, a verification agent independently confirms it, and the change is landed via the project's pull request and merge queue process. ## Quick Start Start the fix-task workflow for this bug and let the investigate, implement, verify, and land nodes run as their prerequisites complete.

Frequently Asked Questions about fix-task

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

FAQPage Schema
How do I automate a bug fix workflow with AI agents?▼

Define the fix as a hub with four nodes: investigate, implement, verify, and land. Each node is a scheduled agent session that fires automatically when its prerequisites complete, starting with the investigate node on approval.

How do workflow nodes know when to start?▼

Nodes declare prerequisites with the relies_on field, such as implement depending on investigate. A node stays blocked until every prerequisite completes, and session outcomes write back to their nodes as annotations.

What happens when a workflow node fails repeatedly?▼

A failing node suspends its own lane after repeated failures and must be re-approved to re-arm. Revoking a node's effect halts that lane while downstream nodes simply remain blocked.

Can I choose which AI agent and model runs each workflow step?▼

Yes, each node's TOML block can specify an agent, model, and effort level, such as claude-code with claude-fable-5 at max effort. Nodes without these settings use defaults.

How is a fix verified before it is merged?▼

The verify node independently re-runs the test battery and, where supported, a live check. It completes only on proof; if verification fails it annotates the failure and does not complete, blocking the land node.