ds-swarm — Claude Code skill: what it does, how to install
Claude Code skill splits broad tasks across parallel workers and merges their verified outputs into a single report. Triggers on /ds-swarm during audits or multi-branch races.
MIT 537 installs ★ 320 updated 2026-10-28
npx skills add aicodeking/deepseek-stack --skill ds-swarm -a claude-codeWhat ds-swarm does
The skill slices a task into distinct partitions or sets up a race where several workers tackle the same problem. Every worker receives a standalone brief specifying targets, exact commit hashes, verification methods, and an isolated output folder. Writable state is never shared between workers.
When an agent harness lacks subagent execution, briefs run sequentially inside fresh contexts. Each worker reports PASS, ISSUES, or BLOCKED backed by command execution logs. Any result missing concrete evidence or referencing wrong commits triggers exactly one respawn; a second failure gets recorded as a gap.
What it helps with
- split codebase audits across parallel workers
- run competitive races between alternative implementations
- aggregate test logs into an evidence-based report
- deduplicate multi-worker bug findings by root cause
Example prompts for the agent
- /ds-swarm audit each module in src/ across separate workers
- Race two refactoring approaches and pick the winner based on benchmarks
- Split this migration across workers and collect proof for every completed slice
How to install ds-swarm
1Skills CLI (Vercel)
npx skills add aicodeking/deepseek-stack --skill ds-swarm -a claude-code2Manually: put the files in place
~/.claude/skills/ds-swarm/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/ds-swarm/
1Skills CLI (Vercel)
npx skills add aicodeking/deepseek-stack --skill ds-swarm -a codex2Manually: put the files in place
.agents/skills/ds-swarm/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add aicodeking/deepseek-stack --skill ds-swarm -a cursor2Manually: put the files in place
.cursor/skills/ds-swarm/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install aicodeking/deepseek-stack skills/ds-swarm2Manually: put the files in place
.github/skills/ds-swarm/SKILL.md1Skills CLI (Vercel)
npx skills add aicodeking/deepseek-stack --skill ds-swarm -a gemini-cli2Platform CLI
gemini skills install https://github.com/aicodeking/deepseek-stack --path skills/ds-swarmManually: put the files in place
.devin/skills/ds-swarm/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add aicodeking/deepseek-stack --skill ds-swarm -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about ds-swarm
How to install ds-swarm in Claude Code?
Run npx skills add aicodeking/deepseek-stack --skill ds-swarm -a claude-code in your terminal.
What does ds-swarm do?
It coordinates parallel workers on task slices or races, enforces command log evidence for every result, and combines outcomes into a deduplicated report.
How does ds-swarm handle worker failures?
A worker that omits terminal evidence or checks the wrong commit gets respawned once. A second failure marks the slice as an unpassed gap.
Source text
SKILL.md · MIT
--- name: ds-swarm description: "Fan out N parallel workers on slices of a job or on a race, collect every result, and return one evidence-backed report. Use for wide coverage, audits, races between approaches or exploration, or for /ds-swarm, 'swarm this', 'split this across workers'." --- # ds-swarm Split work that is wide but shallow across parallel workers, then merge their proven results into one report. ## Steps 1. State the done condition and the report the swarm must return. 2. Pick the shape. Partition the job into non-overlapping slices, or race N workers on the same brief, or both. For a race, decide up front how you will choose: first pass, rank all, or best one. 3. Write one brief per worker that stands alone: goal, exact slice, how to verify, what to report. When workers test specific commits, name the exact commit hashes and the measurement method. 4. Give every worker that writes files its own output location. 5. Spawn all workers at once. If the harness cannot run subagents, run the briefs one by one in fresh contexts. 6. Each worker reports PASS, ISSUES or BLOCKED with evidence: commands run and their output. ISSUES lists every provable issue, not just the first. 7. Collect results. Reject any result with no evidence or the wrong commit and respawn that worker once. A second miss is a recorded gap, not a pass. 8. Merge into one report: the verdict, each slice or arm with its evidence, gaps, and duplicates merged by root cause. ## Stop when - Every slice has a result or a recorded gap and the report is written. - Or a race has a winner under the rule from step 2. - Never respawn a worker more than once. ## Rules - Workers do not share writable state. - A missing slice is never counted as passing. - Report what workers proved, not what they claimed.
© aicodeking · MIT · aicodeking/deepseek-stack