Skip to content
Back to skills

Roast

ASecurity

Roast a startup or business idea. It is funny and savage, but every burn is backed by a real, sourced fact, and it ends with the kill-my-idea verdict card. Use when the user says "roast my idea", "roast my startup", "roast <company>", or wants a shareable, funny verdict. Also works on famous companies (for example Juicero or Quibi).

  • 2 stars
  • 0 votes
  • 1 copy
  • 1 view
  • Added November 11, 2026
ai-agentspythongobash

Security analysis

A100/100

Scanned November 11, 2026

npx -y skills add RemiPelloux/kill-my-idea --skill roast --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Roast?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Roast
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/remipelloux-roast/badge)](https://www.skillsdirectory.com/skills/remipelloux-roast)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: roast
description: Roast a startup or business idea. It is funny and savage, but every burn is backed by a real, sourced fact, and it ends with the kill-my-idea verdict card. Use when the user says "roast my idea", "roast my startup", "roast <company>", or wants a shareable, funny verdict. Also works on famous companies (for example Juicero or Quibi).
---

# Roast My Idea

A roast is the kill-my-idea verdict told as stand-up. It is shareable on purpose.

## Steps

1. **Research and score** exactly as in
   `${CLAUDE_PLUGIN_ROOT}/skills/kill/SKILL.md`, sections 0–5: the safety rules, the brief,
   the kill-shot checklist, the source rules (minimum 10 sources from 6 domains for a roast),
   the metrics, the rubric, the formula and the fixed rules. Famous companies need no
   questions; use public, well-documented facts.
2. **Write the data files** so `compare` can rank the idea: `idea.json` and `verdict.json`
   (`mode: "roast"`) in `ideas/<YYYY-MM>-<slug>/`, following the kill skill's schemas.
   The HTML report is optional for a roast; write it (kill skill §6) if the user asks.
   For a famous company, the roast may use hindsight (that's the joke), but if the user asks
   "would it have worked?", score it as of its launch date (`as_of`, kill skill §3) and keep the
   hindsight in the roast lines only.
3. **Write the roast** to `roast.md` in the same folder, in this order:
   - the **verdict card** (kill skill §6), then the badge;
   - **5 burns**: one line each, one fact each, one source link each. Funny comes from the
     fact itself (numbers, quotes, irony), not from insults. Order them from the biggest
     kill shot down;
   - **"The one thing that's actually good"**: one honest line (the steelman, kill skill §5);
   - **The fix**: the cheapest test or best rescue path that could save it, in at most 2 lines,
     with its projected score;
   - one closing tagline, at most 12 words.
4. **Self-check** with `python3 "${CLAUDE_PLUGIN_ROOT}/scripts/check-idea.py" ideas/<folder>`
   when Bash is available (it accepts a folder without report data).
5. **Answer in chat** with the full roast. It must fit in one screenshot: no extra
   sections, no preamble.

## Rules

- Roast the idea, never the person: no looks, identity or private life.
- Facts only. Every burn has a source, and opinions are labelled as opinions.
  This keeps it funny and fair, and stops it from becoming defamatory.
- Real companies and public figures: only public, documented facts.
- Quotes must be verbatim from the linked source. If you can't confirm the exact words,
  paraphrase without quotation marks.
- Same language as the user.
- Never post anything on the user's behalf.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.