Reviews code against the three qualities of high-quality code in Laminar Software Design: simple to understand and change, correct in behavior, and fast enough for its purpose. A simplicity finding names the cognitive failure a reader meets, a correctness finding names the missing prevention decision or the missing test, and a performance finding names the waste a known alternative avoids. A coding style skill gives the same book's design directions, each with its reason, for any code that is written or changed.
Navigate AI conversation logs like a debugger — discover sessions, drill into turns & tool calls, and extract lessons and workflows from past agent conversations.
Git commit, push, and pull-request workflows with a fixed safety floor: stage by name, refuse staged secrets, and — in the pull-request flow — never put a commit on the remote default branch.
Browser automation for agents on top of the playwright-cli CLI (@playwright/cli): playwright-cli (official skill, vendored), showme (hand the user a headed window in a given state), page-bridge (feedback toolbar on a dev page), pick (one-shot element picker)
KubeRocketCI testbed and workspace toolkit - provision the source workspace and a try-kuberocketci testbed, then diagnose, reproduce, fix, and verify code changes on the cluster, including end-to-end Jira-tracked issue triage, plus opt-in krci-kubelock, which locks kubectl, helm and every script Claude runs to the testbed context
Companion for working on a Lightcone project — an ASTRA analysis (astra.yaml) executed with the `lc` CLI. Covers the project model (a uv project, direct or containerized, stored as a DataLad dataset), the development workflow (allocate a cluster with `lc compute`, probe with `lc run`, integrate recipes, commit, `lc materialize`), recipe resources and GPUs, status interpretation (current/behind/stale/no recipe) and failure diagnosis, publishing the RO-Crate view, and how to engage the user at each phase: interview to scope a brand-new project, orient and summarize when resuming an existing one. Invoke whenever the user wants to start, resume, plan, run, debug, or discuss a Lightcone/ASTRA analysis project. Bundles the `astra` plugin (spec reference, validate-on-save & session-start hooks) — install `lightcone` alone to get the full stack.
File-based project backlog in committed files: open items in docket/DOCKET.md, closed items archived per year with verification records, ids never reused. The record skill captures and closes items, groom sweeps for stale and ghost work and picks what to do next; a dependency-free CLI, bundled MCP server, and self-contained HTML board keep the mechanics byte-exact.
Machine-checked evidence for agents: which registered checks are fresh, stale or failing for the code as it is now, and a gate that is READY only when all of them passed.
Task queue on tasks/*.md where only the verifier marks done, a Stop hook chains tasks, a planner asks without stopping, a conductor runs parallel workers, reviewers and a bug-fix loop, and every task gets the simplest orchestration pattern that is enough.
Route second-tier work to Codex CLI: routing rubric, workflow fan-out worker, computer-use verification. Requires the codex CLI and codex@openai-codex. Vendored unchanged from Millon15/cc-millz.
Roslyn C#/VB language server for Claude Code: one shared per-workspace server, analyzer shadow-copying so builds never lock, native .slnx and .NET 10, plus a refactoring MCP (solution-wide rename, code actions, find-references, format) and an extension API for live systems.