flow

Plugin flow from marketplace ai-plugins in ralphshao/ai-plugins. Commands read from the repo's manifest — 7 days ago.

Skills (14)

• brb

  Keeps this session's prompt cache warm during a break, so the first turn back reads the cache instead of rewriting the whole context. Claude Code only.

• change-body

  The shape of a code review description (pull request, Swarm review, or changelist description) - a one-sentence why, a minimal summary visual, before/after evidence, and a merge-danger call. Use when writing or updating one.

• deep-review

  Multi-pass review of uncommitted work, a branch, or committed code (git SHA or range, PR, Perforce changelist, Swarm review). Runs parallel flow:code-reviewer subagents, one per lens (correctness/spec, standards/quality, history, security when the diff touches a trust boundary, and silent failures w

• flow

  Start or resume an engineering task. Sizes the work (trivial, normal, large) and walks it through start, plan, build, and ship with two human gates.

• host-github

  How flow's review-host operations (fetch-issue, fetch-review, open-draft, ready-for-review) run on GitHub, with gh or the GitHub MCP tools. Load when a flow phase or deep-review names one of these operations and the repo's origin is on github.com.

• host-swarm

  How flow's review-host operations (fetch-review, open-draft, ready-for-review) run on Helix Swarm for a Perforce workspace. Swarm has no issues, so there is no fetch-issue. Load when a flow phase or deep-review names one of these operations and the p4 server has a Swarm URL.

• interview

  Interview the user about a plan, design, or decision in rounds of numbered questions with recommended answers, until no decision is left open. Use when a plan has open decisions, or the user says "interview me", "grill me", or asks to stress-test an idea.

• plan

  Second flow phase. Explores the code, interviews the user on open decisions, writes .flow/<slug>/plan.md, and stops at GATE 1 for approval.

• retro

  Retrospective on a coding session. Suggests fixes to the agent's environment (checks, standards, steering files, tooling) that would have prevented this session's mistakes, most severe first.

• ship

  Last flow phase. Verifies the work, runs deep-review, fixes confirmed findings, strips .flow/, and readies the review for GATE 2.

• start

  First flow phase. Turns a prompt or issue into an isolated workspace (branch, worktree, or changelist) and a checkpointed .flow/<slug>/brief.md.

• tdd

  Test-driven development in red-green slices at agreed seams. Use when building a feature or fixing a bug test-first, or when the user mentions TDD or red-green-refactor.

• vcs-git

  How flow's VCS operations (find-state, isolate, checkpoint, diff-scope, publish, drop-state, land, resolve-target) run in a git repo, on a regular branch or in a worktree. Load when a flow phase or deep-review names one of these operations in a git work tree.

• vcs-perforce

  How flow's VCS operations (find-state, isolate, checkpoint, diff-scope, publish, drop-state, land, resolve-target) run in a Perforce (Helix Core) workspace, with pending changelists or a task/dev stream. Load when a flow phase or deep-review names one of these operations in a p4 workspace.

Subagents (4)

• code-reviewer

  Read-only code reviewer. Scans files or a diff for correctness bugs, silent failures, project-standard violations, untested changes, and improvements to readability, performance, and best practices, showing current code and a fix for each issue. Use when asked to review, audit, or suggest improvemen

• review-validator

  Read-only skeptic that checks code-review findings someone else reported. Tries to refute each one and returns a verdict line per finding (CONFIRMED, REFUTED, or UNSURE). Doesn't look for new issues. Use to validate review findings before reporting them.

• strong-reviewer

  The code-reviewer on a stronger model, for deep-review's correctness and security lenses. Same instructions and limits as code-reviewer. Use only when deep-review asks for it.

• tester

  Test coverage reviewer and writer. Scans files or a diff, maps what existing unit and integration tests cover, finds untested behavior, and writes or updates tests to fill the gaps, then runs them. Use when asked to find missing tests, check test coverage, or add/update tests for code.

Hooks (7)

• PreToolUse

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/guard.py"

• PreToolUse

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/codex_agents.py"

• PostToolUse

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/format.py"

• Stop

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/stop_gate.py"

• Stop

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/brb.py" stop

• UserPromptSubmit

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/brb.py" prompt

• UserPromptExpansion

  "$(command -v python3 || command -v python)" "${CLAUDE_PLUGIN_ROOT}/hooks/brb.py" expand

Similar plugins