Agent Harness Fault Injection
Test how an agent workflow recovers from timeouts, crashes, and lost state using deterministic fault schedules.
- Repository
- sickn33/agentic-awesome-skills
- Stars
- 49,023
- Licence
- MIT
- Files
- 1 file
What it does
This skill teaches an AI agent to deliberately break parts of an agent workflow in a controlled, repeatable way, such as a tool call timing out, a worker restarting, or memory being cleared, and then check whether the workflow recovers without losing state, exceeding its budgets, or repeating side effects. The agent produces a small fault matrix, an event timeline, and a verdict (recovered, contained failure, unrecoverable, or inconclusive). It is meant for testing, not for running on live systems.
Great for
- Engineers building multi-step or multi-agent workflows who need proof that retry, resume, and checkpoint logic works before release
- QA or reliability teams who want reproducible failure evidence instead of random chaos testing
- Developers adding a new recovery path to an agent that calls tools, sandboxes, or memory stores
Skip it if
- Anyone wanting to test a live or production system: the skill explicitly says not to target production, real user data, or live credentials, because injected failures could cause real damage
- Non-technical users without an agent workflow to test, since the output assumes a harness, fixtures, and state machines
Try asking
- “Create a deterministic fault schedule that times out the search tool on the second call and restarts the worker after its second checkpoint, then report whether the task resumes without duplicating side effects.”
- “Build a fault matrix for my multi-agent summarizer covering sandbox denial, MCP timeouts, and parallel branch failure, and give me a verdict for each.”
- “Check whether my agent rejects a corrupted or stale checkpoint on resume instead of guessing progress.”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
agent-harness-fault-injection.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip agent-harness-fault-injection.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip agent-harness-fault-injection.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
- Define the recovery contract and expected terminal states before running any injection, otherwise the verdict will be inconclusive.
- Run the same deterministic schedule twice and compare timelines before trying different schedules or randomized chaos.
- Use synthetic inputs and stubbed tools, and keep network disabled unless a local test server is genuinely needed.
What's inside
- SKILL.md
Redistributed unmodified from sickn33/agentic-awesome-skills under the MIT 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.