timharris707 /
skills
A catalog of 23 agent skills that let one session plan, build and review for you
- 43 stars
- 4 stars since 3 Oct
- Python
- Updated today
- Some experience
- Agent skills & plugins
The author has led dev teams for over twenty years without writing code. Agents kept failing in the same ways: building the wrong thing, giving one model's opinion on a hard decision, writing prose that sounds like AI, and losing context when a session ran long. Each skill is a written-down fix for one of those patterns.
A skill is a plain-markdown SKILL.md file the agent reads on its own. Fifteen are adapted from Matt Pocock's Skills For Real Engineers, three from Lauren Tan's pstack, and five are original. The author rewrote them for someone who won't read the code, then tied them to one orchestrator session. That session takes your idea and decides what gets questioned, mapped, prototyped, built and reviewed.
The repo ships two editions: original Claude Code plugins, and a Codex desktop plugin adapted for Astra workflows. Any runtime that can read markdown can use the skills if you copy or symlink them in.
What it does
Grilling before building
The agent interviews you in rounds about a plan until nothing important is still assumed. It fetches facts itself, and the decisions stay with you.
Advisory board
Claude, Codex, Gemini and Grok review one decision independently, read each other's notes and return one recommendation. It is read-only by default, you approve what leaves your machine, and each run writes a verdict.json you can gate CI on.
Orchestrator and handoff
One session coordinates parallel work lanes. A handoff skill writes a snapshot so a fresh session can resume when context fills up.
Plain writing rules
plainspoken stays on once installed and keeps agent output, commit messages included, free of common AI tells. writing-for-humans and huh cover pages for strangers and messages that didn't land.
Media ingestion and investigation
ingest turns a video or recording into a transcript, frames and manifest. Research, prototype and codebase-review skills answer questions from sources or throwaway code.
What you could build
- Run a setup interview that binds a repo to an orchestrator session you talk to like a dev lead.
- Interrogate a feature plan with grilling before any code is written.
- Put an expensive architecture or launch decision to four models and read one memo back.
- Hand a long task to a fresh session without losing context.
- Turn a video or recording into an evidence packet an agent can reason over.
Get started
In Claude Code, add the marketplace and install the team-workflow plugin, then run the setup skill once in your repo. Open a session and say "orchestration mode". Codex desktop users have a separate install, shown in the README.
/plugin marketplace add timharris707/skills/plugin install team-workflow@skillscodex plugin marketplace add timharris707/skillsCommands are copied from the project's README. Read it on GitHub for the full setup.
More in Agent skills & plugins
superpowers
A set of skills that makes coding agents plan, test and review before they write code
mattpocock/skills
Small, editable agent skills for planning, testing and reviewing code with AI
ECC
Agents, skills, hooks and memory that add an engineering workflow to coding agents
andrej-karpathy-skills
A CLAUDE.md file that tells Claude Code to ask first, keep code small, edit narrowly
anthropics/skills
Anthropic's example Agent Skills: folders of instructions Claude loads per task
ponytail
A skill that makes AI coding agents write the least code that works
gstack
Slash-command skills that give Claude Code roles like reviewer, QA and release engineer
addyosmani/agent-skills
A pack of 25 Markdown skills that guide AI coding agents from spec to ship
taste-skill
Agent skills that steer AI coding tools toward better layout, type and motion
archify
Agent skill that turns a description or a codebase into an interactive HTML diagram