Cascading Failure Detection
Probes an agent's workflow for failures that spread across steps, retries and dependencies.
- Repository
- Tencent/AI-Infra-Guard
- Stars
- 7,320
- Licence
- Apache-2.0
- Files
- 1 file
What it does
This is a security test script for AI agents, not a tool that makes an agent do real work. It gives an agent a set of scenarios, such as a critical tool being down or an early step receiving bad input, and checks whether the agent would let one failure take down the whole chain. Findings are written up in a structured report format with severity and a suggested fix, such as adding circuit breakers or bounded retries.
Great for
- Teams auditing a multi-step agent workflow for fragile dependencies
- Security reviewers running a red-team pass on agent behaviour before release
Skip it if
- Simple, isolated single-step agents, where the skill itself says to skip it
- Anyone expecting it to fix failures automatically; it only probes and reports, and its probes are phrased as text prompts to the agent, so results are judged by the agent's own answers rather than by running the system
Try asking
- “Check whether my agent workflow still completes if the database API is unavailable”
- “Probe my agent for unbounded retry loops when a task is invalid”
- “Test whether my pipeline propagates corrupted data from an early tool call”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.
cascading-failure-detection.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip cascading-failure-detection.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip cascading-failure-detection.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Use it against a realistic copy of your workflow so the scenarios match real dependencies
- Treat the agent's self-reported answers as a starting point and confirm important findings by testing the real system
What's inside
- SKILL.md
Redistributed unmodified from Tencent/AI-Infra-Guard under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.