Skip to content
cccskills
FreePublished on GitHub

grill-me

Interview the user about any topic one question at a time, using the grilling engine with a question set generated from the prompt. Use for "grill me on X" outside of planning. Pass --rounds for batched rounds.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90 stars(GitHub 90 + cccskills 0)

Run score Not tested

See how to install

Files included (2)

  • SKILL.md2.9 KB
  • agents/openai.yaml115 B

SKILL.md (original)

Review the instructions the agent will receive before you install.

/grill-me

Thin entry to the grilling engine (.claude/skills/grilling/SKILL.md). This skill builds the question set; the engine asks it.

Arguments

  • <topic>: what to grill the user on. If empty, ask for it in one AskUserQuestion call.
  • --rounds: pass mode: rounds to the engine. Without it, pass mode: one-at-a-time. Never switch modes on your own.

Procedure

  1. Anchor. Resolve the company: explicit mention in the topic, then the cwd's company folder, then the repo's owner in companies/manifest.yaml, then the session's company_slug. When a company resolves, read its policies folder. If none resolves, run unscoped and read no company files. Set project_dir when the topic or cwd names a project folder of that company; otherwise leave it unset.
  2. Research facts. Read what the topic points to (board, repo, knowledge) so known answers are not asked. Record settled answers as known_facts.
  3. Generate the question set. Write workspace/tmp/grill-me/{date}-{topic-slug}.md in the format in .claude/skills/_shared/questions/FORMAT.md:
    • 4 to 12 questions. At least 4 must be kind: decision.
    • Cover tiers in order: strategic (goal, scope, audience), architecture (approach, boundaries), quality (risks, success measure).
    • Each decision has 2 to 4 short options and a recommended option.
    • Use depends_on where one answer changes the next question. Use kind: fact for anything you can look up.
  4. Validate. bash .claude/skills/grilling/frontier.sh count <set> --known <ids>. On exit 2, fix the set and validate again.
  5. Open a journal when project_dir is set: bash .claude/skills/_shared/journal.sh open grill-me <project_dir>.
  6. Run the engine with question_set, known_facts, and mode. Follow its procedure exactly.
  7. Summarize. Write a short summary: topic, each decision and its answer, assumed answers, open items, and the count object.
    • Always write it to the journal: journal.sh append <project_dir> decisions "summary: ..." when a project journal is open, otherwise a /journal session entry.
    • When project_dir is set, also write it to {project_dir}/grill-{YYYY-MM-DD}.md, then run journal.sh close <project_dir> "<one-line summary>".
  8. Report in chat: the decisions in plain words and where the summary was saved. When the topic is a project, suggest /brainstorm or /prd as the next step.

Rules

  • One AskUserQuestion call per question unless --rounds was passed.
  • Never answer your own decision questions, except under CLAUDE_HEADLESS=1 as the engine defines.
  • Do not read another company's files or credentials.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

accept

Free

Accept HQ membership invites from a magic link or raw token.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90Updated Oct 26, 2026

adr

Free

Capture qualifying architecture decisions as ADRs in the right HQ or repo location.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90Updated Oct 26, 2026

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90Updated Oct 26, 2026

Find shallow modules and propose deepening refactors, ranked with file refs; never edits code. Use when code is hard to change or planning a refactor.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90Updated Oct 26, 2026

Generate dark ASCII banner graphics for HQ and social launch assets.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90Updated Oct 26, 2026

Compare approaches to best-in-class prior art, merge existing work, verify cruft removal, and recommend the most durable design before PRD work.

The English summary is being prepared. Showing the original description.

indigoai-us/hq-core90Updated Oct 26, 2026

See all skills by indigoai-us

Report a problem with this skill