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 cyanseek/dsh-tool-chaos --skill dsh-chaos-testgit clone --depth 1 https://github.com/cyanseek/dsh-tool-chaosWrote 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/cyanseek/dsh-tool-chaos/dsh-chaos-test)<a href="https://agentmods.dev/skills/cyanseek/dsh-tool-chaos/dsh-chaos-test"><img src="https://agentmods.dev/badge/skills/cyanseek/dsh-tool-chaos/dsh-chaos-test/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/cyanseek/dsh-tool-chaos/dsh-chaos-test"><img src="https://agentmods.dev/badge/skills/cyanseek/dsh-tool-chaos/dsh-chaos-test.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00091 | $0.01745 |
| Opus 5 | $0.00046 | $0.00873 |
| Sonnet 5 | $0.00018 | $0.00349 |
| Haiku 4.5 | $0.00009 | $0.00175 |
Grade B, and why
dsh-chaos-test scanned grade B with 1 finding 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 12d 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.
Unrestricted tool accessmediumExcessive agency
A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.
DSH row patches replace the complete `config`; do not write partial overrides. Keep the reporter row enabled in all three overlays. Never broaden to exact `tool: '*'` as a shortcut. How it starts
The opening of the file, as written. The whole thing — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DSH Chaos Test
Turn a resilience requirement into a real, isolated experiment. Do not stop at a plan when the local environment permits execution.
Requires Node.js 22.19+ and a usable dsh CLI. The wrapper can install the published dsh-tool-chaos npm package when it is not already present.
Recognize these requests
- “验证第一次工具失败后 Agent 会不会自动重试。”
- “Prove the fallback works when GitHub/web/database fails once.”
- “测试 timeout 是否真正取消了工作，不能留下孤儿进程。”
- “Verify a blocked tool result is not trusted.”
- “Test a nested Code Mode tool call.”
- “Generate reproducible DSH resilience evidence for CI.”
Do not use this skill for production infrastructure chaos, load testing, penetration testing, or generic security scanning.
Default operating mode
Act autonomously. Resolve repository paths, project instructions, tool names, current Git state, local versions, and existing tests by inspection. Ask at most one concise question only when a fact is both undiscoverable and safety-critical.
Use these defaults unless repository evidence requires a narrower choice:
- current repository as the workspace;
- a disposable profile derived from the workspace path;
- one exact or narrowly globbed tool rule;
probability: 1,afterMatches: 0,every: 1,maxInjections: 1;- baseline first, then
dryRun: true, thendryRun: falseonly after the dry-run proves the intended match; - all artifacts under
.dsh-chaos/runs/<run-id>/; - no push, publish, credential change, permission escalation, production target, or out-of-workspace mutation.
Missing evidence is INCONCLUSIVE, never PASS.
Use the bundled command wrapper
Set <skill-root> to this skill directory. Run:
node <skill-root>/scripts/dsh-chaos.mjs doctor --json
The wrapper prefers an already installed dsh-tool-chaos binary and otherwise uses npx -y dsh-tool-chaos.
Workflow
1. Diagnose and choose the smallest experiment
Read applicable AGENTS.md/contribution rules, inspect the protected behavior and its tests, and run the doctor. Select one primary fault:
What ships with it
6 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.
- 12d ago First seen · 199 lines · 91 tokens per session scan B abfcf6e0427e
dsh-chaos-test is a skill published in the GitHub repository cyanseek/dsh-tool-chaos (4 stars, last pushed 16d ago), licensed MIT. It adds 91 tokens to every session and 1,745 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it B with 1 finding (unrestricted tool access). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
agent-resilience-testing
Orchestrates end-to-end resilience testing for LLM agents with AgentBreak, including LLM infrastructure failures, prompt injection, agent skill supply-chain risk, guardrail verification, and MCP server/tool failures. Use when the user asks to "test my agent for resilience", "chaos test this agent", "find failure modes…
qa
Browser-based QA verification. Launches a real browser, navigates the app, clicks buttons, fills forms, and tests user flows. Works as a standalone skill or as a phase end condition in campaigns. Requires Playwright (optional dependency, graceful skip if not installed).
live-preview
Mid-build visual verification loop. Takes screenshots of components during construction, not just after. Catches visual regressions and invisible features before they compound. Requires Playwright or similar screenshot tool.
test-gen
Generate and verify tests — happy path, edge cases, error paths — using the project's own framework and patterns.
verify
Self-test the Citadel hook pipeline from within a live session. Exercises real tool calls (Write, Edit, Bash, Read) and checks that hooks fired, telemetry accumulated, and no errors occurred. Reports HOOK HEALTH: PASS or HOOK HEALTH: FAIL with a per-hook breakdown.
Autonomous Agent Testing
Testing patterns for autonomous AI coding agents like Devin and SWE-Agent including task verification, output validation, sandboxed execution, regression testing for agent behavior, and safety guardrails for autonomous code generation.