swarm-implement

Orchestrates parallel exploration, scoped implementation, and independent review for multi-file coding tasks.

Updated Sep 14, 2026
One-click install
npx skills add https://github.com/pandejesal/drone-nav-sar --skill swarm-implement-pandejesal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm-implement
Source: https://github.com/pandejesal/drone-nav-sar/tree/main/.swarm/bundled-skills/swarm-implement
Command: npx skills add https://github.com/pandejesal/drone-nav-sar --skill swarm-implement-pandejesal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Complex coding tasks handled by a single-threaded assistant often miss edge cases, skip verification, or break sibling work in shared repositories. This Skill enforces a disciplined swarm workflow so implementation work is explored in parallel, validated objectively, and independently reviewed before completion. ## Core Features & Use Cases - Parallel Exploration: Dispatches subagents for repository mapping, pattern discovery, and side-effect analysis while keeping the main context focused. - Scoped Implementation with Review Gates: Implements in coherent chunks, then requires an independent reviewer and a separate critic to approve the actual diff before work is considered done. - Worktree Safety Discipline: Prohibits destructive git operations in subagents and enforces scope declarations for every delegated file change. - Use Case: When fixing a bug that touches auth logic across several files, use this Skill to explore the codebase in parallel, implement the fix in scoped units, run tests, and get independent reviewer and critic approval before merging. ## Quick Start Ask the assistant to implement your feature or bug fix using the swarm-implement workflow with parallel exploration and independent review.

Frequently Asked Questions about swarm-implement

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a multi-file code change with parallel agents?

Use the swarm-implement workflow: dispatch parallel explorer subagents for repository mapping and pattern discovery, then implement in scoped units with declared file lists. Each change is validated by tests, an independent reviewer, and a final critic before completion.

What is the swarm-implement workflow for code review?

After implementation and objective validation, an independent reviewer subagent inspects the actual current diff and evidence, returning APPROVE, NEEDS_REVISION, or BLOCKED. A separate critic then challenges the approved diff, and any later edit invalidates both approvals.

When should I use the deeper validation path?

Use it for high-risk work: auth, permissions, payments, destructive actions, dependency changes, public APIs, schemas, migrations, concurrency, caching, subprocesses, secrets, and large cross-file refactors with correctness risk.

Can subagents run git stash or git reset during parallel work?

No. Subagents share a worktree with sibling agents, so worktree-global destructive commands like git stash, git reset, or git checkout -- . are prohibited because they destroy sibling agents' in-flight work. The orchestrator handles dirty trees before dispatch.

What happens if a reviewer returns NEEDS_REVISION?

NEEDS_REVISION or BLOCKED blocks completion until the issues are fixed and the change is re-reviewed. Any edit after approval invalidates it, requiring a fresh reviewer pass and another critic pass.