skills.rest
JasonColapietro
JasonColapietroCOMMUNITY
Explore all skills by this author →
Website𝕏@johnnysuedeGitHub

suede-recommend-next-action

Inspects repo, terminal, and plan state read-only to recommend one next action with a runnable prompt.

123|10|Updated May 24, 2026
Search Tags:#workflow#claude code#codex#task planning#recommendation#prompt generation#next action
One-click install
npx skills add https://github.com/JasonColapietro/suede-creator-skills --skill suede-recommend-next-action-jasoncolapietro
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: suede-recommend-next-action
Source: https://github.com/JasonColapietro/suede-creator-skills/tree/main/skills/suede-recommend-next-action
Command: npx skills add https://github.com/JasonColapietro/suede-creator-skills --skill suede-recommend-next-action-jasoncolapietro

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? After a review, audit, plan, or stalled task, it is often unclear which single move matters most. This Skill reads the current repo, terminal, plan, or handoff state and returns one evidence-backed next action packaged as a short copy/paste prompt, instead of another broad audit or vague advice. ## Core Features & Use Cases - Evidence-based selection: Runs read-only checks such as git status, git log, plan file reads, and URL fetches before recommending anything. - Scored candidate ranking: Generates 2-4 candidate actions and scores them on goal alignment, unblocking, evidence, urgency, and leverage. - Three prompt depths: Returns a short runnable prompt by default, with expanded operator and granular step-by-step versions on request. - Use Case: After finishing a code review in a Claude Code or Codex session, ask what to do next and receive one concrete action with a paste-ready prompt naming the target, scope, and done signal. ## Quick Start Ask the agent to inspect the current repo and plan state and recommend the single next move with a short runnable prompt.

Frequently Asked Questions about suede-recommend-next-action

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

FAQPage Schema
How do I get a recommendation for what to do next in my project?▼

Ask the agent what to do next in a Claude Code or Codex session with this skill installed. It reads current git status, recent commits, and any plan or handoff files, then returns one recommended action with a short copy/paste prompt.

How does the skill decide which next action to recommend?▼

It generates 2-4 candidate actions and scores each from 0-2 on goal alignment, unblocking, evidence, urgency, and leverage. Ties are broken by preferring prerequisites, verifiable evidence, and more reversible actions.

Can the next-action recommendation execute the suggested task automatically?▼

No. The skill is strictly read-only while recommending and never mutates files, repos, deployments, or live systems. The recommended action only runs if the user separately authorizes execution.

What is the difference between the quick, expanded, and granular prompts?▼

The quick prompt is a 2-4 sentence runnable instruction returned by default. Saying 'expand prompt' renders a full operator prompt with target, scope, and verification fields, and 'make it granular' decomposes the work into atomic numbered steps with exact commands.

When should I use suede-agent-teams instead of a next-action recommendation?▼

Use suede-agent-teams when the work requires coordinating multiple specialists across parallel lanes. The next-action skill is designed for selecting a single focused move, not multi-lane orchestration.

Related Skills

Explore More Product & Management Skills→
NicktheQuickFTWNicktheQuickFTW

qa-analyst

Audit implementation plans and code changes against a quality rubric.

Community
Basic
RagnarokButMemorySafeRagnarokButMemorySafe

tasks-do

Claim, implement, verify, and update backlog tasks against acceptance criteria.

Community
Advanced
TravisLeeeeeeTravisLeeeeee
110

Orion

Coordinate multi-agent project execution with prioritized tasks and deadline tracking.

Community
Intermediate
dazuckdazuck
1

brainstorming

Generate multiple approaches, analyze tradeoffs, and deliver a final recommendation.

Community
Basic