swarm

Plugin swarm from marketplace startup-swarm in AnimeshShaw/startup-swarm. Commands read from the repo's manifest — 19 days ago.

Skills (19)

• dept-compliance

  Compliance playbook: map target market and data to regimes and obligations. Use when writing a compliance map.

• dept-delivery

  Delivery playbook: backlog, sprints, task files, dependency graph, parallel safety. Use when planning build work.

• dept-devops

  Infrastructure playbook: hosting choice, CI, environments, secrets, deploy, rollback, observability. Use for infra work.

• dept-finance-legal

  Finance and legal playbook: cost model structure, unit economics, licence and contract flags. Use for cost modelling.

• dept-marketing

  Marketing department playbook: positioning, channels, pricing hypothesis, launch plan, metrics. Use for go-to-market work.

• fw-bmc

  Business Model Canvas where every block is a testable hypothesis. Use when shaping a business model from research.

• fw-market-sizing

  Bottom-up TAM/SAM/SOM with visible formulas and ranges. Use when a market size is needed.

• fw-porter

  Porter's Five Forces done with evidence. Use when judging how attractive and defensible a market is.

• fw-pricing

  Pricing hypotheses from value, competitors and unit economics. Use for any price or plan decision.

• method-adr

  Architecture decision records with real alternatives and preserved dissent. Use for any significant technical decision.

• method-blind-audit

  Audit a change from spec, tests, diff and CI result only. Use when reviewing implementation work.

• method-decision-pack

  One-page pack for the human's sign-off at gate G2. Use when compiling the decision pack.

• method-evidence-plan

  Designs the real-world tests that must pass before building. Use when writing an evidence plan.

• method-prd

  PRD with ruthless MVP scope and testable acceptance criteria. Use when writing a product requirements document.

• method-red-team

  Gate-time adversarial review of strategy or design artifacts. Use before G1, G2 and G3.

• method-test-first

  Tests written before code, traced to acceptance criteria, with abuse cases. Use when writing tests or a test plan.

• method-threat-model

  Threat modelling from the design: assets, boundaries, STRIDE threats, concrete mitigations. Use when writing or reviewing a threat model.

• method-verify-sources

  How to check that citations are real and say what is claimed. Use when verifying a document's sources.

• orchestrate

  Run the swarm as orchestrator: dispatch roles, enforce gates, keep the plan. Use after /swarm:run.

Slash commands (8)

• /swarm:approve

  Approve a gate. Usage: approve G1 | approve G1 exploratory | approve G2 | approve G3

• /swarm:end

  End the swarm run and turn its hooks off

• /swarm:god

  God mode, TOKEN-INTENSIVE: never asks you anything, decides from its own research. Usage /swarm:god [full] <problem>

• /swarm:pause

  Pause the swarm so it stops auto-continuing

• /swarm:reject

  Reject a gate with reasons. Usage: reject G2 <reason>

• /swarm:resume

  Resume a paused swarm run

• /swarm:run

  One-shot, TOKEN-INTENSIVE: give a problem, the swarm researches, designs, builds and audits it. Usage /swarm:run [god] [full] <problem>

• /swarm:status

  Show swarm run status and metrics

Subagents (18)

• architect

  Owns technical coherence. Writes ADRs, plans phases and tasks, reviews merges, re-plans stuck tasks. In the lean preset also covers delivery, devops and cost modelling.

• auditor

  Reviews a change against the spec using only the spec, tests, diff and CI result. Structurally cannot read the implementer's notes. In the lean preset it also writes the red-team gate challenges.

• compliance-analyst

  Maps the target market and data types to the regulatory regimes that apply and the obligations they create.

• delivery-manager

  Breaks the approved design into sprints and tasks with a dependency graph; writes the release report with run metrics.

• devops

  Hosting choice, CI pipeline, environments, secrets handling, deploy and rollback.

• engineer-junior

  Implements low-complexity, tightly specified tasks (glue code, CRUD, config, docs). Same rules as the senior engineer.

• engineer-senior

  Implements one task against locked tests. Cannot edit tests or artifacts. Runs CI through the runtime wrapper.

• finance-legal

  Cost model (hosting, APIs, fees, support), unit economics and licence or contract flags. Not legal advice.

• market-researcher

  Answers one research question with sourced notes. Spawned in parallel by the strategy head.

• marketing-head

  Positioning, channels, pricing hypothesis and launch plan, built from the market brief and PRD.

• orchestrator

  Runs the swarm as a thin state machine. Dispatches roles, enforces gates and budgets, batches questions for the human. Never produces artifacts or code.

• product-head

  Owns what gets built. Writes the PRD with testable acceptance criteria and compiles the one-page decision pack for the human.

• qa

  Writes the tests before any code, from the acceptance criteria. Arbitrates test disputes. Tests are locked when QA finishes.

• red-team

  Standing dissenter. Argues against each gate decision before the human sees it; its dissent is logged, never voted away.

• security-engineer

  Specialist for one deep dive (auth flow, data store, a dependency set, a scanner run). Spawned by the security lead.

• security-lead

  Threat model at design time, independent security audit and veto at build time. In the lean preset also maps compliance obligations.

• source-verifier

  Independently re-opens every cited source in the market brief and checks it says what the brief claims. Cannot see the researchers' notes.

• strategy-head

  Turns the brief into a cited market brief and an evidence plan that names real-world tests. Applies strategy frameworks only where evidence supports them.

Hooks (7)

• UserPromptSubmit

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" prompt

• PreToolUse

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" dispatch

• PreToolUse

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" guard

• PostToolUse

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" after-task

• PostToolUse

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" post-write

• Stop

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" stop

• SubagentStop

  sh "${CLAUDE_PLUGIN_ROOT}/hooks/run.sh" substop

Similar plugins