deepseek-and-destroy

deepseek-and-destroy is a skill for Claude Code from mtarcure/claude-vibe-squad. It costs 75 tokens per session (10,160 once invoked), scanned A, original, MIT.

A workflow for carrying out a large, multi-stage coding plan through worker agents, reviews, repairs, and repeated checks. It keeps the main coordinator focused on decisions while delegated agents investigate, implement, verify, and review.

In plain words
What is it for?
Use it for complex implementation plans that span several phases, including codebase surveys, subsystem work, testing, review, recovery after failures, and phase approval.
Why use it?
It reduces the need for one agent to hold every detail of a large repository task at once. It also provides a process for finding and repairing problems before moving to the next phase.

Skill for Claude Code

Written for Claude Code: PreCompact hook event. Also seen: mentions CLAUDE.md; mentions Claude Code; mentions AGENTS.md.

Good fit Use it for complex implementation plans that span several phases, including codebase surveys, subsystem work, testing, review, recovery after failures, and phase approval.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/mtarcure/claude-vibe-squad/dsd
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add mtarcure/claude-vibe-squad --skill dsd
Clone the repo
git clone --depth 1 https://github.com/mtarcure/claude-vibe-squad

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for deepseek-and-destroy

README.md
[![agentmods](https://agentmods.dev/badge/skills/mtarcure/claude-vibe-squad/dsd.svg)](https://agentmods.dev/skills/mtarcure/claude-vibe-squad/dsd)
Your own site
<a href="https://agentmods.dev/skills/mtarcure/claude-vibe-squad/dsd"><img src="https://agentmods.dev/badge/skills/mtarcure/claude-vibe-squad/dsd.svg" alt="Measured on agentmods" height="20"></a>
Per session 75 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 10,160 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00075 $0.10160
Opus 5 $0.00037 $0.05080
Sonnet 5 $0.00015 $0.02032
Haiku 4.5 $0.00007 $0.01016

Measured 3d ago against content hash 9f36c016e965, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

deepseek-and-destroy scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

departments/sysmgmt/reference/dsd-SKILL.md · 920 lines

How it starts

The opening of the file, as written. The whole thing — 920 lines — stays where its author put it; the contents beside it link to each section on GitHub.

DeepSeek and Destroy

Feed it a plan. Keep going until the plan is done—or until a human is genuinely required.

You are the main orchestrator for a complex, multi-phase implementation plan. You own plan-wide understanding, project-aligned judgment, decomposition, routing, escalation decisions, and phase approval. Worker agents perform the tool-heavy and context-heavy work: state surveys, subsystem discovery, implementation, verification, review, repair, recovery audits, and phase evidence synthesis.

The orchestrator owns decisions, not investigation volume.

Do not move repository-scale reading, call-graph tracing, large-artifact analysis, hash collection, repetitive verification, or crash-damage inspection into the main orchestrator merely to improve worker prompts. Express that work as bounded read-only or implementation units, delegate it, then judge the durable evidence. The main orchestrator is not a second implementer, a second reviewer, or a second test runner. It does not modify project source, tests, generated deliverables, or runtime artifacts. It may write orchestration state, decisions, task definitions, and remediation plans.

Worker authority contract

Workers execute and establish technical facts. The orchestrator routes, decides, and approves.

Until a phase gate, all investigation, implementation, verification, review, repair, re-review, recovery analysis, and evidence generation belong to workers or mechanical helpers. The orchestrator must rely on their durable evidence.

If the orchestrator doubts a claim, the doubt is a routing event—not permission to inspect the code, rerun a command, reparse an artifact, or repair the work itself. Commission a fresh worker in a clean context to answer the exact question:

  • a fresh Reviewer for implementation correctness;
  • a Verification Worker for a command, measurement, artifact, or runtime claim;
  • a Discovery Worker for ownership, architecture, or call-path uncertainty;
  • a Recovery Auditor for suspect partial changes;
  • a fresh Phase Auditor for phase-wide evidence synthesis.

Read the full file on GitHub · 920 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 920 lines · 75 tokens per session scan A 9f36c016e965

Subscribe to this mod's changes

deepseek-and-destroy is a skill published in the GitHub repository mtarcure/claude-vibe-squad (109 stars, last pushed 3d ago), licensed MIT. It adds 75 tokens to every session and 10,160 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-05.