squad-mcp

Cradle-to-grave executor for the full squad workflow: brainstorm → grillme → tasks → next → implement → review. Auto-invokes each sub-skill in sequence and halts only at explicit user-decision gates. The user fires ONE command at the start and answers inline gate questions thereafter — no re-typing slash commands between phases. Triggers: "run the full squad pipeline", "take this feature cradle-to-grave", "walk me through from idea to merge", "whole feature flow", "guide me end-to-end", "/squad:pipeline". Pass --from <phase> to enter mid-sequence. Example: "/squad:pipeline new notifications system".

Pre-implementation brainstorming with parallel web research + specialist agents. Use BEFORE coding to decide what to build, compare 3-5 alternatives, scan industry patterns, surface trade-offs, weigh "build vs buy". Produces an options matrix with cited sources and a recommendation — never code, never file edits. Triggers: "brainstorm X", "research approaches to", "explore options for", "compare X vs Y", "should we use X", "help me think through", "what does the industry use", "best practices for", "what's the right approach to". Example: "/brainstorm --deep should we replace our queue layer". For stress-testing a plan you've already drafted, use /squad:grillme. For implementing the decision, use /squad:implement.

Conventional Commits message generator from current staged + unstaged changes. Read-only — runs only allowlisted git commands, never commits, never pushes, never adds AI co-author trailers. Output is plain text the user pastes. Triggers: "suggest a commit", "commit message", "commit msg", "write a commit", "gen conventional commit", "what should this commit say", "/commit". Classifies type (feat/fix/chore/refactor/docs/test/perf); subject under 50 chars; body only when "why" isn't obvious from the diff.

Read-only bug investigation skill. Takes a bug description + optional stack trace + optional repro steps; orients via code-explorer (Haiku grep/excerpt), then dispatches the debugger persona to emit N ranked hypotheses (1/3/5 on --quick/--normal/--deep) with file:line evidence, verification steps, and confidence labels. Never writes code, never commits. Triggers: "investigate this bug", "what could cause X", "help me debug", "why is X failing", "diagnose this error", "root cause of", "trace this issue", "why does Y throw", "intermittent failure". Example: "/squad:debug --normal 500 errors started after rate-limit change". For "where does X live" without a failure yet, use /squad:question. To fix after diagnosis, use /squad:implement.

Opt-in setup for auto-journaling capture plumbing. Copies bundled PostToolUse hook scripts into .squad/hooks/ and prints the .claude/settings.json snippet to wire them up. Capture-only — squad behaviour does NOT change until a follow-up release adds distillation + retrieval. Triggers: "enable journaling", "turn on auto-journaling", "set up the work-trail hook", "capture tool calls", "opt in to journal", "/squad:enable-journaling". One-shot setup — no args, no flags.

Socratic plan validation. Grills your plan one question at a time against the project's domain language (CONTEXT.md) and prior decisions (ADRs in docs/adr/), updating both inline as terms resolve. Catches terminology drift, hidden assumptions, and undocumented decisions BEFORE implementation. WRITES to CONTEXT.md and docs/adr/ with per-write confirmation — use --no-write for dry-run. Triggers: "grill my plan", "stress-test this plan", "validate this approach", "is this consistent with our domain", "are there decisions missing", "before I implement", "challenge my assumptions", "/grillme". Example: "/squad:grillme --deep store user notifications in a separate table". For exploring options BEFORE a plan exists, use /brainstorm. For implementation, use /squad:implement.

Read-only code Q&A. Spawns the code-explorer subagent (Haiku-class) for grep/glob/excerpt and synthesizes a cited answer with file:line citations. Fast. Never writes files, never commits, never runs the squad. Use for "where is X defined", "what calls Y", "how does Z work", "show me the implementation of", "find references to", "explain this code", "locate in the codebase", "trace through", "what files use X", "where does Y happen". Example: "/squad:question where is the rate-limit middleware applied?". For a known failure with stack trace, use /squad:debug instead. For implementing a fix or feature, use /squad:implement.

Observability dashboard for the squad-mcp run journal in this workspace. Reads .squad/runs.jsonl and renders a single-color (cyan) ANSI panel of verdict mix, score buckets, sparkline trend, per-invocation distribution (implement/review/question/debug/brainstorm/grillme), and a per-agent breakdown of token + wall-clock cost. All token figures are estimates (chars ÷ 3.5). Never writes. Triggers: "squad stats", "run history", "score distribution", "where did the tokens go", "how much did squad cost", "agent cost breakdown", "how often does X verdict happen", "/squad:stats". Example: "/squad:stats --quick" (last 7d) or "/squad:stats --thorough --since 2026-04-01".

Multi-agent advisory squad workflow. Two modes — implement (default) runs classify, score risk, select agents, planner, gates, advisory review, implementation, consolidation; review runs advisory-only on an existing diff/branch/PR. Dispatches named subagents (architect, dba, developer, reviewer, security, qa, tech-lead-planner, tech-lead-consolidator, product-owner); each scores its dimension 0-100 and the consolidator weights them into a rubric scorecard. Trigger when the user types /squad:implement, /squad:review, or asks to "run the squad", "advisory review", "implement with squad-dev", "code review by specialists", or invokes any squad-dev workflow.