AGENT SKILL

init-personal-profile

Builds the user's personal profile file config/PROFILE.md through a guided interview and a scaffold script that renders and checks it. Use when a user wants to create or refresh their personal profile, personalize the AI assistant, or types /init-personal-profile. Trigger terms: personal profile, PROFILE.md, onboarding interview, identity, tooling preferences, growth plan, working philosophy. Not for team or role files (init-team, init-expertise) or the agent identity file (init-soul).

License 许可
Apache-2.0
所属仓库
crewrig/crewrig
最近更新
2026-11-15
Stars
★ 24
Forks
6

01 前言 Front Matter

name
init-personal-profile
metadata
{ "provenance": { "version": "1.1.0", "feedback": "https://github.com/crewrig/crewrig", "canonical": "https://github.com/crewrig/crewrig" } }
description
Builds the user's personal profile file config/PROFILE.md through a guided interview and a scaffold script that renders and checks it. Use when a user wants to create or refresh their personal profile, personalize the AI assistant, or types /init-personal-profile. Trigger terms: personal profile, PROFILE.md, onboarding interview, identity, tooling preferences, growth plan, working philosophy. Not for team or role files (init-team, init-expertise) or the agent identity file (init-soul).
allowed-tools
[ "Read", "Write", "Bash", "AskUserQuestion" ]

03 SKILL.md 正文

Help the user write config/PROFILE.md, the file that tells the assistant who they are and how they work. You run the interview, which is judgement; a script renders the file in the template shape and checks it.

Interaction rules

  • Closed questions with a bounded set of choices go through your host CLI's structured-question tool: AskUserQuestion on Claude Code, ask_user on Gemini CLI, ask_question on Antigravity CLI. GitHub Copilot CLI has none, so present the same choices as a labeled plain-text list there. On the other CLIs a plain-text menu is skipped or misread.
  • Open-ended content is free-form chat.
  • A structured question takes 1 to 4 questions of 2 to 4 options; the tool adds "Other" itself, so leave it out of your lists. Keep header labels under 12 characters.

Interview

Ask in the language the user picks; the file itself is written in the language they prefer. Never assume an answer: an invented fact in a profile steers every later session.

  1. Language. Detect the locale with echo $LANG; confirm the conversation language with header: "Language", the detected one first as "(Recommended)", English second.
  2. Identity. Read git config user.name and git config user.email; confirm them (options "Use as is", "Edit name", "Edit email", "Edit both"). Ask Team, Role, Department and Location in chat.
  3. Tooling. Editor and plugins, terminal and shell: chat. Communication channels: multi-select (header: "Channels": Slack, Email, Video call, Chat). Work rhythm (header: "Rhythm": Deep focus blocks, Frequent short bursts, Async-first, Meeting-heavy).
  4. Active projects. Loop: name, responsibility and objective in chat, then header: "Add project" ("Add another", "Done").
  5. Growth plan. Chat: primary focus for the next six months, concrete goals.
  6. Working philosophy. Chat: core values, collaboration preferences, notes. Propose a polished summary and confirm it ("Approve", "Tweak wording", "Rewrite from scratch").

Generation

  1. Write the answers to a scratch JSON file as {identity: {full_name, email, language, team, role, department, location}, tooling: {editor, terminal, communication, rhythm}, projects: [{name, responsibility, objective}], growth: {focus, goals}, philosophy: {values, collaboration, notes}}; leave out what the user skipped.
  2. Run the dry-run command, show data.content, and confirm with header: "Finalize" ("Save as is", "Edit a section", "Discard draft").
  3. Save, run the linter, and close with a short encouraging message.

Scaffold scripts

A command cannot bundle files, so the scripts ship with the init-team skill, installed beside this command on every CLI. Locate them once before writing:

find . ~/.claude ~/.gemini ~/.copilot ~/.agents ~/.github -maxdepth 6 \
  -path '*/skills/init-team/scripts/scaffold-config-file.ts' 2>/dev/null | head -1

If nothing is found, stop and tell the user to install the init-team skill: writing the file by hand would skip the shape check. Below, <scripts> is the directory of that path. Both scripts exit 0 (ok), 1 (finding), 2 (usage error) or 3 (precondition failed), and take --json.

When Command Exit 1 means
To show the draft node <scripts>/scaffold-config-file.ts --kind profile --answers <answers.json> --dry-run (draft in data.content) An answers field is missing or malformed; the finding names it.
To save the same without --dry-run; add --force only when the user chose to replace an existing file target-exists: config/PROFILE.md exists and was not written; ask the user whether to overwrite.
After saving node <scripts>/lint-config-file.ts --kind profile config/PROFILE.md A section is missing, empty or still a placeholder; fix the answers and save again.

The answers file is JSON written to a scratch location; its fields are listed by node <scripts>/scaffold-config-file.ts --help.

04 同仓库更多 Skills