⏳ This skill is pending AI review.
Scores will appear once the review pipeline completes.
dalio-5-step
>-
Choose how to use this skill
You do not need every option. Choose the path your AI client supports. The stable page stays the same; versioned files are immutable.
1. Native installer
This listing has no registered native installer command. Use the complete package or source fallback below, depending on what your client supports.
Do not guess an installer command or replace an existing version without reviewing the diff.
2. Complete package recommended
Download the ZIP when available. It includes SKILL.md plus the references, security notes and version metadata.
No complete ProSkills package is published for this listing yet.3. Prompt-only
Copy the prompt above when the agent can read the stable page or when you want to adopt the workflow without installing a skill.
Need only the instruction file?
Download SKILL.md only if your client requires a single file. The complete ZIP is safer for a full installation because it preserves the references and release context.
No path installs or executes anything by itself. Your agent still needs access to the project files. Before updating, compare the installed version and review the diff.
// RATINGS
Not yet listed on ClawHub or SkillsMP
// README
Skills.md
The Claude Code skills I actually use, kept in one place so other people can steal them.
A skill is a folder with a SKILL.md in it. The frontmatter description is the only
part Claude reads up front — when your request matches it, the rest of the file loads
and Claude follows it instead of improvising. So a skill is really a written-down
procedure: the thing you'd otherwise re-explain every time.
I write one when I notice I've explained the same process to Claude three times, or when the default behaviour is nearly right but skips the step that matters.
What's here
| Skill | Folder | What it does |
|---|---|---|
dalio-5-step | personal | Interviews you through Ray Dalio's 5-Step Process, one step at a time, and won't let you skip to solutions |
people-intelligence | personal | Turns your notes about a person into advice for one specific upcoming conversation |
interactive-thinking | personal | Makes you argue for a position instead of reading an analysis — stakes, competing perspectives, then attacks your strongest reasoning |
plan-setup | work | Points the three skills below at your repo — writes .claude/plan-config.md. Run once per repo |
plan-brief | work | Interviews you into a PRD and writes the ticket — refuses vague answers, one question at a time |
plan-handover | work | Plan a feature with Claude, then hand it to the team as a decision-note PR + per-ticket implementation plans in Jira |
plan-implement | work | Take a ticket that already has those plans and build each sub-ticket to a green PR |
personal/ works for anyone out of the box — no repo, no tooling, no accounts.
work/ is my team's feature workflow — see
the end-to-end walkthrough — and it's generic: /plan-setup writes a
.claude/plan-config.md in your repo with your branch, paths, commands and tracker,
and the other two skills read it. It does assume Jira and the GitHub CLI. Each
folder's README says more.
Install
Skills go in ~/.claude/skills/ (available everywhere) or .claude/skills/ inside a
project (that project only). The skill folder must be a direct child — the
personal/ and work/ split here is for browsing, not for nesting.
git clone [email protected]:derrickleemy/Skills.md.git
cd Skills.md
# one skill, user-wide
cp -R personal/dalio-5-step ~/.claude/skills/
# or symlink, so a git pull updates it
ln -s "$PWD/personal/dalio-5-step" ~/.claude/skills/dalio-5-step
Run /doctor to confirm it was picked up. (On another agent, see
Written for Claude Code below.)
Written for Claude Code
That's what I use, so that's what these assume: a SKILL.md in ~/.claude/skills/,
picked up by description, invoked as /name. Nothing here is deep magic though —
a skill is a procedure written down, and any agent that can be handed a procedure
can follow it. If you're on something else, the substance carries over and you swap
the mechanics.
The Claude-specific pieces, and what they are underneath:
| What the skills use | What it's doing | Swap for |
|---|---|---|
~/.claude/skills/<name>/SKILL.md | how a skill gets discovered and loaded | wherever your harness keeps reusable instructions — a prompt file, a custom mode, a rules directory |
/name invocation | running one deliberately | however you trigger a saved prompt |
/doctor | confirming a skill was picked up | your harness's equivalent, or just ask it to list what it loaded |
plan mode + ExitPlanMode | read-only exploration, then one approval gate before anything is written (/plan-handover §1–2) | any explicit "propose, then wait for my yes" step. The gate matters more than the mechanism — without it the skill writes files before you've agreed to the direction |
/goal | re-running a turn until a condition holds, which is what carries /plan-implement across five sub-tickets unattended | a loop or eval harness if you have one; otherwise type "next ticket" between them and babysit. The skill says this and degrades cleanly |
/code-review high | a second pass over the diff with fresh context (/plan-implement §5) | any review command, or a subagent told to review the working diff |
/mcp | connecting the Atlassian connector | however you attach tools; the Jira calls themselves are MCP, not Claude |
~/.claude/plans/<generated-name>.md | where a plan-mode plan gets parked before the skill copies it somewhere permanent | wherever your harness writes its scratch plans — or delete the step and write the plan directly |
| the scratchpad path for PR bodies | keeping a body file outside the repo so it can't land in a commit | any temp directory. Don't skip it — this one bites |
The Jira tool calls (getJiraIssue, editJiraIssue, createJiraIssue, …) come
from the Atlassian MCP server, so they work with any MCP-capable agent. gh and
git are just CLIs.
work/ skills also carry Jira and GitHub assumptions beyond the harness —
work/README.md covers those.
Tools I use alongside these
Skills cover the procedure. These cover the parts of the loop Claude Code isn't:
| Tool | What it does | Where it fits |
|---|---|---|
| crit | Review plans, diffs and live pages in a browser UI, leave inline comments, send them straight back to the agent | The human review gate — instead of reading a plan in the terminal and typing paragraphs of feedback, comment on the line you mean. Pairs with /plan-handover's approval step |
| NotebookLM | Upload sources, ask questions grounded only in them, audio overviews | Reading-in before a plan exists — dump the RFCs, vendor docs and meeting transcripts in, interrogate them there, then bring the conclusions into /plan-brief |
Use
Claude picks a skill up on its own when your request matches the description — that's the point, and it's why the descriptions are long and full of the phrases I actually type. You can also invoke one directly:
/dalio-5-step
/people-intelligence how should I approach Sam about the deadline slipping
Three things that make skills work better, learned the hard way:
- Be specific in the request. "Prep me for the 1:1 with Sam about the promotion he didn't get" beats "help with Sam" — the skill has more to work with, and a vague ask is what makes Claude guess.
- Answer the questions properly.
dalio-5-stepand anything else interview-shaped is only as good as your answers. One-word replies get a one-word-quality worksheet. - Edit the skill when it's wrong. These are text files, not a product. If the procedure doesn't fit, change the file — that's cheaper than arguing with Claude about it every session.
Why I bother
Claude is good at the work and bad at my order of operations. Left alone it jumps to a plan before the problem is named, writes a technical doc when a colleague needed prose, or gives me a personality profile when I asked how to open a hard conversation. A skill fixes that once instead of every time — and because it's a file, the process is now reviewable, diffable and shareable, which the version living in my head never was.
License
MIT — take them, fork them, make them yours.
// HOW IT'S BUILT
KEY FILES