mpreibisch-claude

Plugin mpreibisch-claude from marketplace mpreibisch-skills in Michael-Preibisch/mpreibisch-skills. Commands read from the repo's manifest — 13 days ago.

Skills (9)

• adversarial-review

  Ask the other model to challenge a plan, implementation or PR, or idea with evidence-based findings ranked by severity. Use for a second opinion, to challenge a plan, to find holes in a design or diff, or for any critical second-model review without code changes.

• artifact-standards

  House standards for every shareable HTML page an agent produces - artifacts, visual explainers, reports, plans, recaps, comparisons, architecture explainers, and any page meant to be shared or posted in team chat. Load it before you write or update such a page, including a quick one-pager. It fixes

• brainstorm

  Brainstorm with the other model. Both generate ideas independently, then build on and challenge each other's ideas, and rank them against the user's criteria while keeping dissent. Use to brainstorm together, generate alternatives, or explore options with Codex or Claude Code; not for reaching agree

• clear-writing

  Make all written output clear, concise, and human-readable - documentation, code comments, chat replies, summaries, explanations, commit messages, and rewrites of existing text. Based on the Google developer documentation style guide, ASD-STE100 Simplified Technical English, and anti-slop editing ru

• consensus

  Reach a reasoned Codex and Claude Code consensus on a question, plan, decision, or diff through independent assessments and bounded disagreement resolution. Use to cross-check a decision with the other model, resolve a disagreement between the models, or get an explicit shared conclusion or split de

• decisions

  Use the OpenAI Decisions API for fixed alternatives over sufficient supplied text when the judgment replaces useful work. Prefer it to spawning a sub-agent or calling a small LLM just to pick a label, status, yes/no, or score. Covers loop checks (is a watched build, job, PR, or agent progressing, st

• delegate-explore

  Delegate a bounded research or codebase-exploration lane to the other model, with an explicit question, scope, and evidence-backed artifact. Use to ask the other model to investigate, trace, map, or compare something, without implementation changes.

• delegate-implement

  Delegate a concrete, testable implementation specification to the other model with bounded write scope, acceptance criteria, and a verified handoff. Use to have the other CLI agent implement an authorized, scoped specification.

• pr-review-ask

  Draft a concise request to review one or more GitHub PRs, explaining why the changes matter and preserving confirmed stack relationships. Use for "ask for a review", "draft a review request", or "these PRs are ready for review". Show the draft in chat; adapt it to a destination only when requested.