flow

Plugin flow from marketplace maude in 1aGh/maude. Commands read from the repo's manifest — today.

Skills (24)

• source-command-bug-fix

  Implement and verify the fix specified by a ticket root-cause analysis.

• source-command-bug-rca

  Investigate a ticket and document its root cause with evidence.

• source-command-done

  Validate and close a feature: decision capture, delivery, retrospective and archive within authorized scope.

• source-command-execute

  Execute an implementation plan and maintain verification and workflow checkpoints.

• source-command-init

  Scaffold the .ai workspace, detect the stack, check instruction files and dependencies; --context, --map and --kgai add project context, an architecture map and kgai import.

• source-command-plan

  Create a feature implementation plan grounded in project requirements and the design system.

• source-command-record-ddr

  Record an architectural or product decision with its rationale.

• source-command-release

  Follow the project release runbook and its action approval gates.

• source-command-run

  Run a plan unattended as a goal with a capped Stop gate; also where you are (--status), pause and resume work, with or without a run.

• source-command-validate

  Run quality gates and validation lanes: full pipeline, --fast, --security, --a11y, --visual, --review, --scenario.

• agent-browser

  Automate websites with agent-browser: navigate, inspect, fill forms, capture screenshots and verify web flows.

• agent-device

  Automate native mobile apps with agent-device: simulator setup, selectors, screenshots and scenario execution.

• context-budget

  Keep agent contexts small: slice briefs, range reads, per-agent caps, hand-backs with Decisions and Continue, lead checkpoints.

• debate-protocol

  Run planning, research or review debates with native subagents, bounded rosters and one consolidated decision.

• debugging-rules

  Debug from evidence and root cause through hypothesis, fix and verification; stop repeated speculative fixes.

• kgai-backend

  Resolve kgai availability and store scope; read prior decisions, capture actual changes and sync when configured.

• orbit-backend

  Resolve or create the orbit task for the current work, report flow working state, and store plans, RCAs and reports in orbit instead of .ai/ — warn-only.

• quality-gates

  Resolve project quality commands, scoped checks, gate failures and validation fallbacks from workflow config.

• question-protocol

  Batch necessary questions from workflow participants, collect user answers and route them to the requester.

• scenario

  Author or run cross-platform UI scenarios with selectors, screenshots, platform coverage and result reports.

• security-rules

  Apply application and AI security checks to code reviews, implementation and threat-driven validation.

• testing-rules

  Apply test quality rules during implementation and review: meaningful assertions, isolation and coverage.

• ui-rules

  UI hard-stop rules and audit method: WCAG 2.1 AA accessibility, responsive layout and motion.

• workflow-state

  Read and update workflow state when planning, executing, completing, pausing or resuming project work.

Slash commands (10)

• /flow:bug-fix

  Implement and verify the fix specified by a ticket root-cause analysis.

• /flow:bug-rca

  Investigate a ticket and document its root cause with evidence.

• /flow:done

  Validate and close a feature: decision capture, delivery, retrospective and archive within authorized scope.

• /flow:execute

  Execute an implementation plan and maintain verification and workflow checkpoints.

• /flow:init

  Scaffold the .ai workspace, detect the stack, check instruction files and dependencies; --context, --map and --kgai add project context, an architecture map and kgai import.

• /flow:plan

  Create a feature implementation plan grounded in project requirements and the design system.

• /flow:record-ddr

  Record an architectural or product decision with its rationale.

• /flow:release

  Follow the project release runbook and its action approval gates.

• /flow:run

  Run a plan unattended as a goal with a capped Stop gate; also where you are (--status), pause and resume work, with or without a run.

• /flow:validate

  Run quality gates and validation lanes: full pipeline, --fast, --security, --a11y, --visual, --review, --scenario.

Subagents (16)

• _security-regex-catalog
• a11y-auditor

  Use proactively after UI changes (/flow:validate) to audit accessibility: keyboard, ARIA, contrast, focus, screen readers. Live axe-core via agent-browser. Reports only.

• breaker

  Internal — spawned only by flow:debate-protocol; never auto-delegated. Dissent seat: argues the strongest case a direction is wrong.

• builder

  Internal — spawned only by flow:debate-protocol; never auto-delegated. Divergent seat: argues the most ambitious viable approach.

• design-system-guard

  Use during /flow:validate on UI changes to check compliance with the project design system: screenshots plus static grep for color, icon and type drift. Reports only.

• ethical-hacker

  Use proactively for adversarial security review of any change, incl. AI/MCP attack surface (prompt injection, tool poisoning). Runs with security-auditor in /flow:validate. Reports only.

• investigator

  Internal — spawned only by flow:debate-protocol and research orchestrators; never auto-delegated. Holds one candidate cause and gathers evidence.

• scenario-runner

  Use when /flow:validate or /flow:done need cross-platform UI verification: runs agent-browser and agent-device scenarios, writes a screenshot report. Does not edit code.

• scout

  Read-only fact finding for a flow plan or run: where X is used, inventories, history mining, checking hand-back claims. Returns a cited table; never edits.

• security-auditor

  Use proactively after changes touching auth, input handling, storage, network IO, dependencies, MCP or prompts (/flow:validate). OWASP-class static scan; reports only.

• shipper

  Internal — spawned only by flow:debate-protocol; never auto-delegated. Divergent seat: argues what survives scope, effort and the existing system.

• spike

  Resolves one open design or architecture question of a flow plan with prototypes and measurements; writes the note workers code against. Spawned by /flow:execute and /flow:run.

• test-coverage

  Use after /execute or before /done to audit test coverage of the changed code: untested paths, missing edge cases. Suggests tests; does not write them.

• tracker-reader

  Internal — spawned only by the orbit-backend regression step (/flow:done 1b); never auto-delegated. Reads tracker plans, returns grammar-bound tokens.

• user-advocate

  Internal — spawned only by flow:debate-protocol; never auto-delegated. Argues who a decision serves, confuses or excludes.

• worker

  Implements one bounded task of a flow plan with fail-first tests in the files the brief names; hands back small commits. Spawned by /flow:execute and /flow:run.

Hooks (5)

• SessionStart

  PKG="${CLAUDE_PLUGIN_ROOT}/../.."; if [ -f "$PKG/cli/lib/preflight.mjs" ]; then (cd "$PKG" && node cli/lib/preflight.mjs --plugin flow --warn-only); elif command -v maude >/dev/null 2>&1; then maude p

• SessionStart

  command -v kg >/dev/null 2>&1 || exit 0; PKG="${CLAUDE_PLUGIN_ROOT}/../.."; if [ -f "$PKG/cli/bin/maude.mjs" ]; then (cd "$PKG" && node cli/bin/maude.mjs kg session-sync --warn-only >/dev/null 2>&1 &)

• SubagentStart

  node "${CLAUDE_PLUGIN_ROOT}/hooks/run-hook.mjs" subagent-start

• SubagentStop

  node "${CLAUDE_PLUGIN_ROOT}/hooks/run-hook.mjs" subagent-stop

• Stop

  node "${CLAUDE_PLUGIN_ROOT}/hooks/run-hook.mjs" stop

Similar plugins