Add Sabot to LLM & MCP Testing

[Jott2121]

Adding Sabot under LLM & MCP Testing, placed alphabetically between mcp-server-fuzzer and Tenro.

What it does that the neighbouring entries don't: Tenro and mcp-server-fuzzer exercise agents from the outside. Sabot injects one controlled fault inside a running multi-agent pipeline (a corrupted tool result, a falsified success report, an altered inter-agent message, a silent model downgrade, stale context, a silent no-op) and scores whether the pipeline's own reviewer and guardrail stages catch it — no external monitor in the detection path.

Why it may be useful to this list's readers: measured across LangGraph, CrewAI and AutoGen/Magentic-One, the median own-check detection rate was 16.7%, and the most common outcome was a correct final answer with nothing ever flagging the fault.

• Apache-2.0 harness, CC BY 4.0 spec and results
• Specification, seeds and adjudication anchors registered publicly before any scored run
• Full raw trace corpus published, so every score is recomputable offline with stdlib Python and no API key: python3 scripts/score_wave2.py --seeds-file ../seeds/wave2.json reproduces all 189 published rows byte-identically
• Archived with a DOI: https://doi.org/10.5281/zenodo.21539796

Disclosure: I'm the author. You merged my other project (Crucible) into Mutation Testing a couple weeks ago, happy to adjust the wording or drop this if it's out of scope for the section.

[semanticdiff-com]

Review changes with  

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[mergify]

Tick the box to add this pull request to the merge queue (same as @mergifyio queue).

• Queue this pull request

[llamapreview]

LlamaPReview — No blocking issues found

Review details and evidence

LlamaPReview checks

• Reviewed changed regions in README.md.
• Read the complete PR-head file README.md.

---

Automated review by LlamaPReview · Free for public open-source projects.

[Jott2121]

Note on the two red checks — I looked into both and neither is caused by this change, which is a single added line:

• pre-commit.ci: the md-dead-link-check hook failed with Temporary failure in name resolution on essentially every link in the README, not just the one I added, the runner had no DNS. The standalone md-dead-link-check GitHub Action, which did have network, passed. The same check is also failing on Add Crucible to Mutation Testing #96, [pre-commit.ci] pre-commit autoupdate #98, Add greenproof to Tools #99 and Add pytest-deltatest to Speed section #100, so it looks environmental rather than PR-specific.
• build: peter-evans/create-pull-request@v3 errored with When the repository is checked out on a commit instead of a branch, the 'base' input must be supplied. That's the workflow running against a fork checkout; it also failed this way on [pre-commit.ci] pre-commit autoupdate #98.

rumdl, typos, md-dead-link-check (the Action) and the whitespace/EOF hooks all passed. Happy to rebase if that would help clear them.