workflow

Plugin workflow from marketplace my-skills in Shearerbeard/claude-skills. Commands read from the repo's manifest — 26 days ago.

Skills (10)

• codex-cli

  Drive the codex CLI from an external harness (Claude Code, OpenCode, a script) for a second-model read. Use when the user says "codex review", "second opinion via codex", "adversarial review", "drive codex", "codex exec", or names codex as the reviewer for a diff, plan, or card. Contains the sandbox

• collaborating-with-antigravity

  Delegate plan, prose, spec, architecture, product, and marketing review to the Google Antigravity (agy) CLI, through the agy MCP tools from Claude Code or a JSON-bridge script from other harnesses. Also covers sandboxed execution in a worktree and long agent loops that should not block the main conv

• gate-probes

  Use before git commit, gh pr create, or when the user says "ready to commit", "review this diff", "check before I push", or "run the gates". Also use when the user says "ensure proper gates", "verifiable checks", "ready for implementation", "plan with gates", or "scope this change". Run these univer

• git-commit

  Use when writing a commit message or preparing a git commit. Covers conventional commit format, message structure, prose quality, and mandatory user review before committing. If a github-workflow skill is available, pair with it for branch naming and PR conventions.

• opencode-cli

  Drive OpenCode from an external harness (Claude Code, codex, a script) through the opencode CLI, for code writing and second-model code review. Use when the user says "delegate to opencode", "opencode review", "second-model code review", "run opencode", "opencode CLI", or names an opencode reviewer

• plan-discipline

  Your planning discipline FOR plan mode: the user's required data-gathering depth, staging, and review gates. Load it whenever you plan - it layers on plan mode, it does NOT replace it. In Claude Code especially, EnterPlanMode is the container and this skill is the contents; entering plan mode withou

• process-feedback

  Use when the user says "file process feedback", "record process friction", "send feedback to boardkit", or names friction whose fix belongs in an upstream toolkit or process repo (boardkit, rust-holes, another kit this repo consumes) rather than in the current repo. Records the finding as a session-

• prose-corpus

  Use when the user says "prose feedback", "capture this prose", "corpus this", "save this as a prose sample", or flags a draft as too verbose, overly prescriptive, or AI-sounding without asking for a rewrite. Also self-triggers before humanizer materially rewrites external prose (a commit message, ti

• rebase-hygiene

  Use only when the user explicitly invokes it - never auto-load. The user's required rebase ritual for long-lived branches and worktrees: fetch before trusting any local state, dry-run the rebase and pre-vet the conflicting surface area, resolve conflicts per pre-planned resolutions rather than ad ho

• skill-retro

  Use when the user asks for a skill retro, says "retro time", "retro my skills", "vet my skills", or asks how skills performed or triggered this session. This reviews the skills, not the work: do not use for code review, sprint or project retros, or postmortems of the work itself.