skills.rest
ETdoFresh
ETdoFreshCOMMUNITY
Explore all skills by this author →
GitHub

do-task

Orchestrates plan, implement, and evaluate phases with automatic retry until tasks pass verification.

1|Updated Feb 6, 2026
Search Tags:#workflow#planning#verification#task automation#retry logic#code implementation#subagent orchestration
One-click install
npx skills add https://github.com/ETdoFresh/claude-marketplace --skill do-task-etdofresh
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: do-task
Source: https://github.com/ETdoFresh/claude-marketplace/tree/main/plugins/do-task/skills/do-task
Command: npx skills add https://github.com/ETdoFresh/claude-marketplace --skill do-task-etdofresh

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Executing multi-step coding tasks often fails silently or produces unverified results. This Skill runs a structured plan-implement-evaluate loop with automatic retries until the implementation passes a runnable verification. ## Core Features & Use Cases - Three-Phase Workflow: Delegates to dedicated planner, implementer, and evaluator subagents, each running in its own context. - Automatic Retry Logic: On a FAIL evaluation, feeds the feedback back to the implementer and loops until the result passes. - Organized Artifacts: Stores the request, plan, implementation summaries, and evaluation reports under a numbered task-results/<slug>/ directory. - Use Case: Ask it to add user authentication to your app; it plans the changes, implements them, runs a verification check, and retries automatically if the check fails. ## Quick Start Run the do-task skill with a task description such as "add input validation to the signup form" and it will plan, implement, and verify the change automatically.

Frequently Asked Questions about do-task

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

FAQPage Schema
How do I run a coding task with automatic verification?▼

Invoke the do-task skill with a task description as the argument. It creates a numbered task folder, delegates planning, implementation, and evaluation to subagents, and retries automatically until the evaluation reports PASS.

How does the retry logic work when implementation fails?▼

When the evaluator's report shows FAIL in its Result section, the attempt counter increments and the implementation phase reruns with the previous evaluation feedback included in the prompt. The loop continues until the evaluation passes.

What files does the task workflow generate?▼

Each run creates a task-results/<slug>/ directory containing request.md, plan.md, implementation summaries, and evaluation summaries. Retry attempts produce numbered files like implementation-2-summary.md and evaluation-2-summary.md.

Does the evaluation require actual test execution?▼

Yes, the evaluator subagent must produce a runnable verification, not just a subjective code review. The evaluation report must include a Result section with either PASS or FAIL.

What are the limitations of this task execution approach?▼

Each subagent runs in its own context without access to the main conversation, so all necessary information must be passed in delegation prompts. Tasks with ambiguous requirements may loop through retries if evaluation criteria are unclear.

Related Skills

Explore More Software Engineering Skills→
EtroxTaranEtroxTaran

plan

Generate task breakdowns with acceptance criteria, file scope, and test plans.

Community
Basic
mathiasbourgoinmathiasbourgoin
2

git-conventions

Standardize git commit messages, branch names, and pull request descriptions.

Community
Intermediate
alangebalangeb

code-review-workflow

Run pyscan, pyanalyze, ruff, and black to generate a unified markdown report.

Community
Intermediate
yusaduymazyusaduymaz

gsd-fast

Execute trivial one-step tasks inline without planning artifacts.

Community
Basic