SkillContext Brain Setup
Turn a short structured interview into a project's context brain: AGENTS.md, CONTEXT.md, and a VOICE.md profile, with a completeness score, a gap list, and a refresh diff when answers change.
Works with: Claude Code, Codex, Cursor, Muse · Version 1.0.0 · Needs: An interview brief (brief.txt: [business], [voice], [working], [decisions] sections, key: value lines).
Get this Skill with Premium See plans
Preview
Sample output
An example of what this Skill produces. Sample content, shown for illustration.
Context Brain Setup report — sample Completeness: 100/100 (9/9 required answers). Sections: business 3/3 — complete; voice 2/2 — complete; working 2/2 — complete; decisions 2/2 rules — complete. Gaps: none. AGENTS.md carries: Project: BeanLoop — a weekly habit tracker for indie founders; Audience: indie founders shipping solo products; Goal: ship v1 in 6 weeks without scope creep; Stack: Python + SQLite + plain HTML; Non-goals: No mobile app in v1. Decision rules on file: scope — If a feature does not serve the weekly review, it waits until v2. evidence — Never claim a number that is not in the inputs. VOICE.md carries: Tone: plain, specific, no hype; Avoid: hype adjectives, vague superlatives; Example (verbatim): "Shipped the weekly review on Friday. 214 users tried it." Refresh (v2 brief): Changed: business.goal, decisions.evidence; Added: decisions.pricing; Removed: none.
What you get
- AGENTS.md — the agent's working memory: business, voice, working style, decision rules
- CONTEXT.md — the project snapshot: facts on file, agreements, and open questions
- VOICE.md — the voice profile: tone, avoid list, and one verbatim example line
- A completeness report plus a refresh diff when answers change (changed / added / removed)
About this Skill
Reads a plain-text interview brief in four sections — business, voice, working style, decision rules — and generates AGENTS.md (working memory), CONTEXT.md (facts, agreements, open questions), and VOICE.md (tone, avoid list, one verbatim example line), plus a completeness report (required answers present / 9 × 100) and a stored brief for later diffs. A rerun with --previous adds a refresh report listing changed, added, and removed answers. Missing answers render as 'Not yet defined' and land in the gap list — nothing is invented, and the script never writes into your project: you copy the three files in by hand. Local-only, standard library only.
Try it
Set up the context brain for my project from this interview brief.Good fit
- Starting a project with an AI agent that needs shared context
- Stopping an agent from guessing facts the project never stated
- Keeping project context current as goals and rules change
Not for
- Writing your answers for you — gaps stay marked, never invented
- Judging whether answers are good (the score measures presence, not quality)
- Auto-installing files into your project — you copy the three files in by hand
How to install
- Copy the install prompt into your agent:
Read https://skillu.dev/install/agent.md, then install the Context Brain Setup skill (context-brain-setup). - Or download the Skill package from this page after launch and place its folder in your agent’s skills directory.
- Call it with a slash command, e.g.
/context-brain-setup, and describe what you want in plain words.
FAQ
What if my brief is incomplete?
You still get all three files, but missing answers render as 'Not yet defined' and appear in the gap list. Fill the gaps in the brief and rerun — never let a context file guess.
Does it install the files into my project?
No. The script only writes to the output directory you choose. Copying the files into the project is a deliberate step, so a half-answered brain never lands silently.