AlexanderMattTurner
Official@AlexanderMattTurner
AlexanderMattTurner publishes skills for sandbox security evaluation, CI failure triage, pull request lifecycle management, and rigorous test engineering.
Agent Skills by AlexanderMattTurner
Showing 16 vetted skills indexed across 1 GitHub repositories.
AlexanderMattTurnerrun-control-tower
Dispatch and debug Control Tower marginal-defense evals on sbx microVMs via GitHub Actions.
AlexanderMattTurnermarkdown-block
Wraps markdown content in a copyable fenced code block with correct backtick escaping.
AlexanderMattTurnerstress-test
Generates smoke and stress tests for sandbox runtimes, firewalls, and infrastructure components.
AlexanderMattTurnerci-triage
Diagnose failing GitHub CI checks and drive root-cause fixes to green.
AlexanderMattTurnerdefect-to-guard
Turns a fixed defect class into a costed guard proposal for CI checks and lints.
AlexanderMattTurnerhandoff
Writes self-contained end-of-session handoffs documenting unresolved failures after all reachable fixes.
AlexanderMattTurnerparallel-audit
Audits a codebase across multiple quality dimensions using parallel read-only subagents and produces a vetted fix plan.
AlexanderMattTurnerrun-ctf
Runs and reports the live-fire breakout CTF that tests sandbox containment against autonomous agents.
AlexanderMattTurnergit-workflow
Guides committing, pushing, and resolving merge conflicts in this repository.
AlexanderMattTurnerdoctrine-capture
Captures user-stated NEVER and ALWAYS rules as committed doctrine in the same session.
AlexanderMattTurnervuln-research
Find, reproduce, and report security vulnerabilities against a stated threat model.
AlexanderMattTurnerexplore-plan
Drives a written Explore-Plan-Critique-Review-Verify workflow for non-trivial coding tasks.
AlexanderMattTurnerwriting-tests
Guides writing, fixing, and reviewing tests that verify real behavior instead of source text.
AlexanderMattTurnerbabysit-prs
Drives a set of open GitHub pull requests to landed via auto-merge and merge-queue monitoring.
AlexanderMattTurnerpeer-review
Reviews pending code changes via a read-only reviewer subagent and fixes findings iteratively.
AlexanderMattTurnerpr-creation
Creates and updates GitHub pull requests with an iterative compress-critique-fix review loop.
Frequently Asked Questions About AlexanderMattTurner
FAQPage SchemaWhat tasks can I accomplish with AlexanderMattTurner's skills?▼
You can run Control Tower/LinuxArena marginal-defense evals and breakout CTF containment tests, triage failing CI checks, create and babysit pull requests through auto-merge, resolve merge conflicts, write non-vacuous tests, convert defects into lint or CI guards, run parallel codebase audits, and capture session doctrine and handoffs.
Who are these skills designed for?▼
Engineers maintaining the glovebox sandbox security stack and its repository. The skills target practitioners running microVM isolation evals, reviewing diffs via read-only reviewer subagents, managing GitHub check state, and enforcing test-quality rules like red-on-old/green-on-new non-vacuity and mutation coverage floors.
How do the security evaluation skills work in practice?▼
run-control-tower dispatches the live sbx-microVM end-to-end eval, builds the cells JSON, selects the model and budget, and pulls .eval transcripts. run-ctf dispatches containment rounds across postures like guarded and stock, then reports verdicts in a results table with explicit limits on what each verdict claims.
What prerequisites or dependencies do these skills assume?▼
They assume the glovebox repository environment with its sandbox runtime, GitHub access via gh locally or mcp__github__* on the web, pytest-xdist and c8/kcov coverage tooling, the ci-truth-serum shared guard pack, and the project's code-reviewer subagent. Vulnerability work requires the classical analysis tools already wired into the tree.
How do these skills handle failing tests and defects?▼
ci-triage treats every red check as an undiagnosed bug requiring log evidence before any re-run. defect-to-guard converts fixed defect classes into proposed lints, pre-commit hooks, or CI checks dogfooded against the real tree, while writing-tests enforces behavior-driven tests proven capable of failing.