Supplies the accepted task format GODS (Goal, Overview, Definition of Done, Solution). Use when the user asks what that format is, or to write up decided work in it. NOT for planning before coding — weighing variants belongs to the planning skill.
66
83%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Passed
No findings from the security scan
The framework helps to formulate tasks in a way that avoids misunderstandings between the "client" and the "executor":
Structure:
G — Goal: Why are we performing the task? What is the business goal?
O — Overview: What is happening now? Why did the task arise? What is happening around it?
D — Definition of Done: When do we consider the task completed? By what criteria?
S — Solution: How can the task be solved?
Moved here from the project instruction template on 2026-08-27. It used to sit
in AGENTS.md, which every session mounts, so the model already had the format
and never opened this skill: write-gods-tasks-trigger-pos-1 scored 0/3 with
13, 16 and 27 tool calls, the agent writing the file itself each time. This
file is now the single source, and AGENTS.md points here.
Path: resolve the tasks role from the project instructions. The template
default is a date hierarchy — <YYYY>/<MM>/<slug>.md, kebab-case slug, max
40 chars — but the project instructions are authoritative.
---
implements:
- FR-XXX
---
# [Task Title]
## Goal
[Why? Business value.]
## Overview
### Context
[Full problematics, pain points, operational environment, constraints, tech debt, external URLs, @-refs to relevant files/docs.]
### Current State
[Technical description of existing system/code relevant to task.]
### Constraints
[Hard limits, anti-patterns, requirements (e.g., "Must use Deno", "No external libs").]
## Definition of Done
Every DoD item MUST pair with (a) an FR-ID and (b) a runnable acceptance reference. Items without this tuple are wishes, not contracts.
- [ ] FR-XXX: <observable behavior>
- Test: `<path/to/test>::<test_name>` (or `Benchmark: <scenario-id>`)
- Evidence: `<command that passes iff the item is done>`
- [ ] FR-YYY: <observable behavior>
- Test: `...`
- Evidence: `...`
## Solution
[Actionable step-by-step approach. Never left blank — a task file without a
Solution is not written yet. When a planning workflow weighed variants and the
user picked one, this covers that variant only; when no variant selection took
place, it covers the approach you are proposing.]Frontmatter: date (YYYY-MM-DD; required), status: to do | in progress | done | superseded (required), implements: [FR-...] (optional — present for FR-driven
tasks, omitted for internal/maintenance), optional tags, optional
related_tasks, optional migrated_from, optional superseded_by (required
when status: superseded). Never edit status by hand mid-flight — commit
workflows derive it from the ## Definition of Done checkbox count.
korchasa/flowai-plugins
240d92b
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.