agent-glovebox

Creates high-quality pull requests with an iterative compress-critique-fix loop before submission. Activate this skill whenever you are asked to create, open, submit, or push a pull request, OR whenever a new feature, fix, or refactor is complete and ready to ship. Also activate when the user says "make a PR", "open a PR", "submit this for review", "push and create a PR", "I'm done, create the PR", "the feature is done", "I'm finished", or any variation of completing work / requesting a pull request. Always activate before running `gh pr create`.

How to run the external Control Tower / LinuxArena eval (glovebox's marginal-defense benchmark) — dispatching the live sbx-microVM e2e or the bare-vs-guarded delta pilot, building the cells JSON, picking the agent model + budget, running concurrent sweeps on separate refs, and pulling the .eval transcripts back. Activate whenever the user asks to "run CT", "run the Control Tower eval", "smoke test the CT harness / all CT environments", "run a guarded-condition eval", "dispatch ct-inspect-e2e", "measure marginal defense / ΔS / ΔU", or wants to scan CT eval transcripts. Consolidates the CT knowledge scattered across CLAUDE.md, tests/eval/control_tower/, and the ct-*.yaml workflows.

How to write, change, or review tests in this repo — the load-bearing rule is test real behavior, not source text. Also covers non-vacuity (red-on-old → green-on-new), covering enumerated sets member-by-member, fuzzing parser/Unicode logic, hostile-pre-state filesystem tests, and the c8 / kcov / mutation coverage floors. Activate whenever the user asks to write, add, fix, refactor, strengthen, or review tests ("write a test", "add tests", "test this", "regression test", "cover this", "why didn't this test catch it"), or when a coding task's last step is testing the change you just made.

Drives the Explore -> Plan -> Critique -> Review -> Verify discipline for non-trivial, multi-file work before any code is written. Activate when the user asks to "plan this", "scope this out", "figure out how to do X", "explore the codebase first", or is starting a change that touches several files or an unfamiliar area. Enforces a self-critiqued written plan and real verification instead of trusting a success claim.

Runs a broad, multi-dimension codebase audit by fanning out parallel read-only subagents (one per dimension x file-area), independently confirms their findings, then groups the survivors into a parallelizable multi-PR fix plan and self-critiques it before delivery. Activate when the user asks to "find dozens of issues", "audit the codebase", "find security/ robustness/testing/UX deficiencies", "use subagents to audit and confirm", "find bugs and missed opportunities", or "make a plan to fix all of these in parallel across multiple PRs". Distinct from peer-review (which reviews a pending diff) and security-review (single-pass, current branch): this audits the WHOLE codebase along several axes at once and ends in a vetted fix plan, not edits.

Generates comprehensive smoke and stress tests for infrastructure components (sandbox runtimes, container orchestration, firewall rules, etc.), prioritizing end-to-end realism that catches real setup issues across platforms. Activate when the user asks to "stress test", "smoke test", "test the sandbox", "verify isolation", or wants comprehensive infrastructure testing for a component.

Updates an existing pull request with new changes, commits, and optionally revises the PR description. Activate when the user asks to update, fix, or add to an existing PR. Also activate when the user says "update the PR", "fix the PR", "add this to the PR", or any variation of modifying an existing pull request.

Runs a fresh, unbiased review of pending changes by delegating to the read-only code-reviewer subagent, then triages and fixes the findings. Activate when the user asks to "review this", "peer review", "get a second opinion", "review my changes/diff/branch", or wants a review pass before opening a PR. Distinct from the built-in /code-review and /review skills: this one drives the project's code-reviewer agent to a fixed point.