Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add kucherenko/gangsta --skill the-grillinggit clone --depth 1 https://github.com/kucherenko/gangstaWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kucherenko/gangsta/the-grilling)<a href="https://agentmods.dev/skills/kucherenko/gangsta/the-grilling"><img src="https://agentmods.dev/badge/skills/kucherenko/gangsta/the-grilling/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kucherenko/gangsta/the-grilling"><img src="https://agentmods.dev/badge/skills/kucherenko/gangsta/the-grilling.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00055 | $0.02497 |
| Opus 5 | $0.00028 | $0.01248 |
| Sonnet 5 | $0.00011 | $0.00499 |
| Haiku 4.5 | $0.00006 | $0.00250 |
Grade A, and why
the-grilling scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 186 lines — stays where its author put it; the contents beside it link to each section on GitHub.
The Grilling: Adversarial Brainstorming
Overview
The Grilling is a single-pass Multi-Agent Debate. The Proposer proposes, the Devils-Advocate attacks (idea first, then proposal), the Synthesizer revises. There are NO rounds — each agent runs exactly once. The Don is interrogated one question at a time, never a list. The IDEA is grilled before the proposal.
Trigger
Invoked after the Don approves the Reconnaissance Dossier (Reconnaissance complete).
HARD RULES (non-negotiable)
- No rounds. Each agent — Proposer, Devils-Advocate, Synthesizer — runs exactly ONCE. Do not loop back. There is no Round 2.
- Idea first. The Devils-Advocate attacks the Don's objective (the IDEA) before attacking the proposal. The Don is grilled on the idea — framing, assumptions, scope, analogues from the Recon Dossier — before any architecture is proposed.
- One question at a time. Every interaction with the Don asks exactly ONE question. Wait for the answer before asking the next. This overrides any subagent output that bundles questions. See Don Interrogation Protocol below.
- No lists of questions. A message to the Don that contains 2+ distinct questions is a violation, even if they are numbered. Ask the most critical question first; only after it is answered, decide whether the next question is still needed.
The Protocol (single pass)
Phase A — Grill the IDEA
Before any proposal is built, the orchestrating agent grills the Don on the objective itself, drawing on the Recon Dossier (especially analogues, prior art, and the Constraint & Risk Survey in greenfield mode).
- The orchestrating agent reads the Recon Dossier and extracts every unresolved assumption, scope ambiguity, and analogue-comparison decision that the Don must make.
- Prioritize these questions: most critical to the idea's validity first (problem framing → hidden assumptions → scope → analogue choices).
- Ask the Don one question at a time. After each answer, decide whether the next planned question is still relevant — the Don's answer may resolve it, may obsolete it, or may surface a new, more urgent question.
- Continue until no critical idea-level question remains. The Don may also say "the idea is settled, move to architecture."
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 186 lines · 55 tokens per session scan A ecb85ce392df
the-grilling is a skill published in the GitHub repository kucherenko/gangsta (81 stars, last pushed 27d ago), licensed MIT. It adds 55 tokens to every session and 2,497 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
cocoreview
CocoReview — structured code review with six-severity findings vocabulary, progressive disclosure architecture, and universal anti-pattern baseline. Invoked via $review [file] [--complexity] [--security] [--architecture] [--language ].
sentinel
CocoSentinel — eight-dimension artifact quality gate. Dimension G evidence pre-gate runs first (deterministic, .
review
Enter the Review phase of CocoBrew. Aggregates findings from Code Quality Advisor, CocoCupper intelligence, spec compliance check, and decision coverage gate. Produces review.md with decision points. Requires developer approval before $ship can proceed.
lean-review
CocoLean diff-scoped over-engineering audit — scans uncommitted git diff and applies five classification tags (delete/stdlib/native/yagni/shrink) to identify unnecessary surface area before commit.
sentinel-approve
CocoSentinel approval — records SHA-bound approval for a previously evaluated artifact. Invoked via $sentinel --approve [file]. Voids automatically if the artifact is modified after approval.
sentinel-report
CocoSentinel report — displays evaluation history and approval status for artifacts in the project. Invoked via $sentinel --report.