rudhra-swarm
Orchestrates multiple agents in parallel to handle complex coding tasks effectively.
Install
mkdir -p .claude/skills/rudhra-swarm && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16720" && unzip -o skill.zip -d .claude/skills/rudhra-swarm && rm skill.zipInstalls to .claude/skills/rudhra-swarm
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Orchestrate multiple specialized agents or agent roles for a project. Use when the user says "rudhra swarm", "spin multiple agents", "orchestrate agents", "parallel agents", "agent team", "multi-agent review", or asks Rudhra to split work across mapper, researcher, implementer, tester, reviewer, and learning roles.Key capabilities
- →Coordinate multiple specialized agent roles for a project
- →Map repository structure, configurations, and risk areas
- →Research current guidance from official documentation
- →Implement minimal code, documentation, or configuration changes
- →Review final diffs for bugs, regressions, and risks
How it works
The skill orchestrates specialized agents (e.g., orchestrator, repo-mapper, implementer) based on a chosen swarm size, running roles in parallel or sequentially, integrating their results, and capturing lessons.
Inputs & outputs
When to use rudhra-swarm
- →Multi-agent review
- →Orchestrating parallel development
- →Splitting complex projects
About this skill
Rudhra Swarm
Use this skill when one agent should coordinate several focused roles.
Default Roles
Use only the roles needed for the task:
| Role | Job |
|---|---|
orchestrator | Owns the plan, integration, final decisions, and user communication |
repo-mapper | Reads repo structure, key configs, commands, architecture, and risk areas |
docs-researcher | Checks official docs or source-backed references for current guidance |
implementer | Makes the smallest complete code/doc/config changes |
test-runner | Runs targeted checks, summarizes failures, and avoids noisy logs |
reviewer | Reviews final diff for bugs, regressions, missing tests, and risk |
learning-capturer | Drafts reusable lessons for agent/PROMOTE.md; never applies them silently |
Workflow
- Read
AGENTS.md,CLAUDE.md,agent/PROJECT_CONTEXT.md, andagent/WORKFLOW.md. - Choose a swarm size:
small: orchestrator + repo-mapper + reviewermedium: small + implementer + test-runnerlarge: medium + docs-researcher + learning-capturer
- State the role split briefly before starting substantial work.
- Run roles in parallel when the environment supports subagents. If real subagents are unavailable, simulate the roles sequentially with clear headings.
- Integrate results in the orchestrator role. Do not let subagents independently commit, push, deploy, or edit the master template.
- End with a learning capture pass:
- project-specific lesson -> project
AGENTS.md,CLAUDE.md, oragent/PROJECT_CONTEXT.md - reusable lesson -> draft row for
agent/PROMOTE.md - master-template change -> only via
harvest-template-lessonsafter approval
- project-specific lesson -> project
Approval Rules
- Ask before deployment, DNS, production data changes, or destructive commands.
- Ask before adding dependencies.
- Ask before promoting lessons into the master template.
- Do not auto-commit or auto-push unless the user explicitly asks.
Output Shape
Keep the final response short:
- roles used
- changes made
- checks run
- learnings drafted for approval
When not to use it
- →When only one agent is needed for a task
- →When the user does not want to split work across specialized roles
- →When not needing to integrate results from multiple roles
Limitations
- →Do not let subagents independently commit, push, deploy, or edit the master template
- →Ask before deployment, DNS, production data changes, or destructive commands
- →Ask before adding dependencies
How it compares
This workflow enables a structured, multi-agent approach to complex projects, allowing for parallel execution and specialized expertise, which is more efficient than a single agent handling all aspects.
Compared to similar skills
rudhra-swarm side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| rudhra-swarm (this skill) | 0 | 1mo | No flags | Advanced |
| workflow-router | 1 | 7mo | Review | Intermediate |
| loki-mode | 1 | 6mo | Review | Advanced |
| wg | 0 | 3mo | Review | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
workflow-router
parcadei
Goal-based workflow orchestration - routes tasks to specialist agents based on user goals
loki-mode
davila7
Multi-agent autonomous startup system for Claude Code. Triggers on "Loki Mode". Orchestrates 100+ specialized agents across engineering, QA, DevOps, security, data/ML, business operations, marketing, HR, and customer success. Takes PRD to fully deployed, revenue-generating product with zero human intervention. Features Task tool for subagent dispatch, parallel code review with 3 specialized reviewers, severity-based issue triage, distributed task queue with dead letter handling, automatic deployment to cloud providers, A/B testing, customer feedback loops, incident response, circuit breakers, and self-healing. Handles rate limits via distributed state checkpoints and auto-resume with exponential backoff. Requires --dangerously-skip-permissions flag.
wg
graphwork
Use this skill for task coordination with WG. Triggers include "wg", task graphs, multi-step projects, tracking dependencies, coordinating agents, or when you see a .wg directory.
run-tasks
JeremyKalmus
Orchestrate task execution via beads and sub-agents. Gets ready work from beads, spawns appropriate agents based on labels, monitors completion, and updates status. Use after /approve-spec has created tasks.
start-dev-team
trysmr
>
sprint-checkin
twardcox
Start of sprint: carry-over, goals, blockers, sprint plan artifact; get_slash_command: sprint-checkin