conon.ai

conon.ai/Tips

Let It Interview You Into A Spec

Don't write the big spec — be interviewed out of it. Have the agent grill you one question at a time, each with its own recommended answer ('your call' is a valid reply), appending every decision to spec.md. Then take that file to a FRESH session: the interview clutter stays behind, the decisions come along.

AdvancedPROMPTINGClaude Code

Why it works

You don't write the big spec — you get interviewed out of it. The agent asks one question at a time, proposes its own recommended answer alongside each (so you can just say 'your call'), and appends every decision to spec.md. Then the crucial move: take the spec to a FRESH session, because the interview's forty turns of deliberation are now noise and the spec is the signal. An hour of this outperforms a day of supervising a build aimed at the wrong target.

How to do it

  1. Start from your raw idea — one messy paragraph is enough.
  2. Run the interview prompt (snippet below).
  3. Answer each question; delegate freely ('your call') on details you don't care about.
    claude code
    Q4: Should bookings require payment upfront?
    My recommendation: no — capture card, charge after
    the session. Lower drop-off for a new service.

you: your call

✓ spec.md updated (7 decisions)
    One question, a recommended answer, and the spec grows behind you.
  4. Watch spec.md grow after every answer — that file is the deliverable.
  5. Interview complete? Read spec.md once end-to-end; fix anything that reads wrong.
  6. Open a NEW session, attach @spec.md, and say: 'build this, plan mode first.'
  7. Keep spec.md updated as decisions change — it outlives every session.

Copy this

prompt
I want to build [X]. Interview me about it relentlessly until we reach shared understanding. One question at a time. With every question, propose your recommended answer so I can accept it fast. After each of my answers, update spec.md with all decisions so far.

What to watch for

  • Building in the interview session drags the whole deliberation along as context — the fresh-session handoff is the point, not a nicety.
  • Answering every question yourself defeats the recommended-answer mechanic; the interview should feel like approving drafts, not sitting an exam.

Sources

Original: conon.aiDocumentation: code.claude.com

Read next

Scheduled Agents Need An Open Lid
Local scheduled tasks run only while the app is open and the machine is awake — a missed slot is skipped, not postponed. Turn on keep-awake, leave the laptop lid open, and write drift-tolerant instructions: 'if it's already past 5pm, skip today and report that you skipped'. Need lid-closed reliability? That's a cloud task — which can't see your local files.
Claude Code
Typed It Three Times? Make A Skill
The third time you type the same procedure is the signal: say 'create a skill that does this' and the agent writes the skill file itself. Then improve it in a loop — run it, note what's off, have it edit its own skill. House rules: 2nd repeated mistake → CLAUDE.md line; 3rd repeated prompt → skill.
Claude Code
Writer Session, Reviewer Session
Run two parallel sessions on one version-controlled project: one builds, the other — with fresh context — reviews. The docs are explicit about why it works: a fresh session isn't attached to code it didn't write, so it critiques honestly where the author session defends.
Claude Code
Send Noisy Research To A Subagent
Any side-investigation that would flood your session — 'which of these fifty files mentions the old pricing?' — goes to a subagent. It reads everything inside its own separate memory and returns three lines; your session's context stays clean for the actual work.
Claude Code

We build and operate the systems behind tips like this one — agents, automations and internal tools, shipped into production. See what the studio does →