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 agentmods add skills/jpantsjoha/ai-native-developer-experience/adversarial-gatenpx skills add jpantsjoha/ai-native-developer-experience --skill adversarial-gategit clone --depth 1 https://github.com/jpantsjoha/ai-native-developer-experienceWrote 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/jpantsjoha/ai-native-developer-experience/adversarial-gate)<a href="https://agentmods.dev/skills/jpantsjoha/ai-native-developer-experience/adversarial-gate"><img src="https://agentmods.dev/badge/skills/jpantsjoha/ai-native-developer-experience/adversarial-gate.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00046 | $0.00843 |
| Opus 5 | $0.00023 | $0.00421 |
| Sonnet 5 | $0.00009 | $0.00169 |
| Haiku 4.5 | $0.00005 | $0.00084 |
Grade A, and why
adversarial-gate 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 5d 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 — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Adversarial Gate
Coined by Jaroslav Pantsjoha (#HarnessEngineering). The Adversarial Gate is the practice of forcing an agent — or a human — to argue against their own design before proceeding. It pre-empts bad reasoning before bad output exists.
The gate has one question: "How would I break this?"
Answer it honestly. If you cannot name at least two plausible failure modes, you do not understand the system well enough to ship it.
When to use
- Before finalising any architecture or design decision
- Before a deployment or release
- When an agent is about to take an irreversible action
- When a design looks clean and elegant (danger signal — simple-looking systems hide the hard failure modes)
- When the team is under time pressure and "just shipping it" feels compelling
Procedure
- State the proposal clearly — in one or two sentences, what is the approach being validated?
- Run the adversarial pass — argue against it. Ask:
- What is the single most likely way this fails in production?
- What happens under load / at scale / with bad input?
- What is the blast radius if this fails? Is it recoverable?
- What assumption does this design make that could be wrong?
- What does the monitoring not cover that would let this fail silently?
- Name at least two concrete failure modes — not "something could go wrong" but specific, nameable failures.
- Check each failure mode has a mitigation or an accepted risk owner — unmitigated = not shippable.
- Document the outcome — pass (risks named and owned) or fail (proceed to redesign). Attach to the ADR or PR.
Outputs
- Adversarial Gate verdict: PASS / FAIL
- Named failure modes with mitigations or risk-owner sign-off
- Optional: a short paragraph appended to the ADR or PR description summarising the gate result
Guardrails
- One argument is not enough. A single failure mode is the one you were already thinking about. The Adversarial Gate is looking for the one you were not.
- "We have tests" is not a mitigation. Tests cover known paths. The gate is looking for unknown paths.
- Time pressure does not suspend the gate. Pressure is when you need it most.
- The gate is not pessimism. It is the fastest route to confidence — because you have already stress-tested the design yourself.
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.
- 5d ago First seen · 66 lines · 46 tokens per session scan A 609c9fb6431d
adversarial-gate is a skill published in the GitHub repository jpantsjoha/ai-native-developer-experience (11 stars, last pushed 29d ago), licensed Apache-2.0. It adds 46 tokens to every session and 843 once invoked, about $0.0002 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
mcp-builder
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
last30Days
Resolve "last30Days" to a concrete ISO date range relative to your run time — a rolling 30-day window ending today. Returns inclusive civil dates plus exact UTC instants so you have temporal context without computing dates by hand. Read-only: no writes, no network. Use before a "last 30 days" / trailing-month task…
thisWeek
Resolve "thisWeek" to a concrete ISO date range relative to your run time — this week so far (Monday → today). Returns inclusive civil dates plus exact UTC instants so you have temporal context without computing dates by hand. Read-only: no writes, no network. Use before a week-to-date task (this week's activity…
comet-hotfix
Comet 预设 —— 通过 open-build-verify-archive 短流程修复已有行为 bug。.
comet-verify
Comet Phase 4: Verify and Close. Invoke with /comet-verify. Verify implementation matches design, handle development branch.
comet-design
Comet Classic 阶段 2 —— 为 change 产出深度技术 Design Doc。.