Menu
Use when creating or updating skills.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Use when writing or modifying ZSH functions.
Use when user says "ralph <dir>" or "ralph this". Implements the highest-priority issue from state.json in the given directory using TDD, updates docs, runs review in a subagent, fixes actionable feedback, then stops for user to commit. Argument is the directory containing state.json and GUIDANCE.md.
Use when user says "sidequest". Compact conversation context into a document for a fresh agent to pick up.
Use when writing or modifying JavaScript code. Apply when adding functions, fixing bugs, or implementing features.
Use when writing or modifying Python code. Apply when adding functions, fixing bugs, or implementing features.
Use when user says "phone pickup", "pick up the conversation from the phone", "récupère la discussion du téléphone", or wants to retrieve a mobile vocal session saved in Notion.
| name | skill-writer |
| description | Use when creating or updating skills. |
Creates and updates skills that make agents follow a process.
Skills are tracked inside the git root at tools/ai/claude/config/skills/<name>/SKILL.md.
Goal: Reach shared understanding of what the skill should enforce.
Exit criterion: You can answer: what behavior should the agent follow, and in what situations?
Use the /grill-me skill to ask one question at a time, about what the skill
should enforce (not how to encode it).
Goal: Produce a skill file that captures the agreed behavior.
Exit criterion: Skill file written at the canonical location.
Read references/skill-template.md.
Create or update the skill file at the canonical location: tools/ai/claude/config/skills/<name>/SKILL.md.
Skill is written. Let the user commit.
| Rationalization | Reality |
|---|---|
"~/.claude/skills/<name>/SKILL.md is where the skill lives, I'll edit it there" | That path is a symlink to the main branch, not the current worktree. Edit inside the git root. |
tools/ai/claude/config/skills/<name>/SKILL.md)references/skill-template.md (Goal + Exit criterion per step, Checklist, Rationalizations)