agent-skills

Plugin agent-skills from marketplace mikhin in mikhin/claude-plugins. Commands read from the repo's manifest — 9 days ago.

Skills (6)

• codex-code-review

  Code review by OpenAI Codex, judged by Claude. A second opinion on a branch diff, the working tree, a commit or a plan from a model that did not write it: Codex reads and reports, Claude verifies every finding against the code, then edits. One or two rounds, Codex never writes. Cross-model AI code r

• effort-advisor

  Recommend an effort level (low, medium, high, max) for a coding task in one or two lines before starting it. Triggers when the user describes a coding task to do — a feature, a bug, a refactor, a review — or asks "what effort", "which effort level".

• multi-agent-debate

  Multi-agent debate until consensus: three agents get contradicting hypotheses and must refute each other with evidence, the session judges, exit on agreement, not on round count. Adversarial review, red team and critique for contested decisions where a mistake is expensive: how many of what is neede

• multi-agent-deep-research

  Deep research by four parallel agents: investigator, analyst, critic and devil''s advocate each research the topic from their angle, then one synthesised report with agreements, tensions, risks and next steps. Multi-agent research for any topic that deserves more than one perspective. Triggers on "d

• overbuild

  Review a diff for code that handles cases no user reaches or builds layers nobody needs — guards, queues, refs, counters, outputs nothing reads. Run on the branch diff as part of every review, and on every review round. Triggers — "overbuild", "/overbuild", "is this overbuilt", "what in this diff is

• verify

  Verify an implementation at high effort — reproduce first, adversarial self-review, randomized tests against a reference, tests that fail on a reverted fix, large inputs — and report what was checked, what failed, what was fixed.

Similar plugins