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/mnvsk97/agentbreak/agent-resilience-testingnpx skills add mnvsk97/agentbreak --skill agent-resilience-testinggit clone --depth 1 https://github.com/mnvsk97/agentbreakWrote 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/mnvsk97/agentbreak/agent-resilience-testing)<a href="https://agentmods.dev/skills/mnvsk97/agentbreak/agent-resilience-testing"><img src="https://agentmods.dev/badge/skills/mnvsk97/agentbreak/agent-resilience-testing.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 | $0.00097 | $0.01089 |
| Opus 5 | $0.00048 | $0.00544 |
| Sonnet 5 | $0.00019 | $0.00218 |
| Haiku 4.5 | $0.00010 | $0.00109 |
Grade A, and why
agent-resilience-testing 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 — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Resilience Testing
This skill is the orchestrator. AgentBreak is the deterministic engine for live fault injection; static checks and guardrail probes fill the gaps that a proxy cannot observe directly.
Core Rules
- Run the real
agentbreakCLI for initialization, validation, inspection, proxying, scorecards, and history. - Cover all four lanes: LLM failure modes, agent skill supply chain, guardrails, and MCP server/tool behavior.
- Treat mock-mode direct requests as endpoint smoke tests, not full agent resilience evidence.
- Treat proxy-mode traffic through the user's actual agent as the primary resilience signal.
- If MCP is enabled, run
agentbreak inspectbefore testing. If inspect fails, stop and ask whether to fix MCP config or disable MCP testing. - If proxy auth or connection validation fails, stop and fix that first. Do not run chaos tests against broken upstream auth.
- Always clean up local processes and restore temporary
.envchanges after a run. - Ask before changing project config files unless the user has already explicitly asked you to perform setup.
Orchestrator Workflow
-
Check AgentBreak
- Run
agentbreak --help. - If missing, detect
uvor an active virtualenv and install with the least invasive command.
- Run
-
Map The System
- Inspect the repo for provider, framework, MCP usage, agent skills/capabilities, guardrails, tool calls, base URLs, retry logic, timeout handling, and error paths.
- Read
references/detection.mdfor scan patterns and what to report. - Present concise findings and choose mock or proxy mode.
-
Build The Four-Lane Test Plan
- Read
references/coverage.md. - Lane 1: LLM failures: outages, brownouts, rate limits, context/token failures, malformed outputs, prompt injection, and streaming failures.
- Lane 2: Agent skills: scan installed/local skills for unsafe instructions, secret access, network exfiltration, downloads, dangerous commands, and hidden prompt injection.
- Lane 3: Guardrails: verify input, output, tool-argument, tool-result, PII/secrets, policy, and human-approval checks.
- Lane 4: MCP: test server failures, tool failures, bad schemas, bad results, registry drift, rug pulls, poisoning, and cross-tool attacks.
- Read
What ships with it
7 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.
- 5d ago First seen · 78 lines · 97 tokens per session scan A 2921f2844597
agent-resilience-testing is a skill published in the GitHub repository mnvsk97/agentbreak (20 stars, last pushed 3mo ago), licensed MIT. It adds 97 tokens to every session and 1,089 once invoked, about $0.0005 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
Angry User Simulator
Simulate aggressive user behavior patterns including rapid clicking, random navigation, form abuse, tab spamming, and unexpected interaction sequences to find UI resilience issues.
executing-distributed-system-tests
Use when running a previously designed distributed-systems test plan against a real or simulated cluster — driving fault injection, workload, chaos scenarios, linearizability / consistency runs, durability, partition, crash-recovery, upgrade, performance/SLO runs, tenant isolation runs, boundary or authz runs…
AI System Quality Engineer
Test LLM, RAG, MCP, and agentic systems end to end. Build golden datasets, run deterministic checks and LLM judges, score retrieval, probe prompt injection, verify tool use, and gate CI on thresholds. Orchestrates DeepEval, Ragas, promptfoo, and Langfuse.
telegram-harness
Run leashd's live tmux+Telegram verification harness in scripts/harness/ — a fake Telegram Bot API plus a real Engine/TmuxAgent/MultiConnector wired like runmulti. Inject messages, slash commands, and inline-button taps and observe the exact outbound streaming/approval/plan/task timeline. Use to reproduce or verify…
reticle
Reticle embeds a dev-only SDK in the user's running app and exposes it to you as reticle MCP tools. You look, act, observe, and assert against the real app. No screenshots, and no browser download for the verify loop: it drives the tab the user already has open.
reticle
Install, instrument and verify this running web app from the inside (DOM, network, routing, console and framework state) instead of screenshots or guessing. Drives one real flow end to end and returns a verdict with the file:line to fix. Use when the user asks to set up or install Reticle, when a user-facing change…