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.
npx skills add bestdeejay-design/agent-skills --skill dsh-runnergit clone --depth 1 https://github.com/bestdeejay-design/agent-skillsWrote 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.
[](https://agentmods.dev/skills/bestdeejay-design/agent-skills/dsh-runner)<a href="https://agentmods.dev/skills/bestdeejay-design/agent-skills/dsh-runner"><img src="https://agentmods.dev/badge/skills/bestdeejay-design/agent-skills/dsh-runner/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/bestdeejay-design/agent-skills/dsh-runner"><img src="https://agentmods.dev/badge/skills/bestdeejay-design/agent-skills/dsh-runner.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00260 | $0.01882 |
| Opus 5 | $0.00130 | $0.00941 |
| Sonnet 5 | $0.00052 | $0.00376 |
| Haiku 4.5 | $0.00026 | $0.00188 |
Grade A, and why
dsh-runner 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 9d 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.
How it starts
The opening of the file, as written. The whole thing — 130 lines — stays where its author put it; the contents beside it link to each section on GitHub.
dsh-runner
Автономный агент DeepSeek Harness (
dsh) как внешний исполнитель задач: изолированная папка-workspace, полный JSONL-лог каждого шага, никакого доступа к файлам вне задачи.
Загружай этот скилл, когда нужно запустить агента на реальной задаче в отдельном окружении: починить баг в клоне репозитория, сгенерировать код по описанию, прогнать одну задачу на разных моделях и сравнить результат.
🎯 When to use
Use this skill when:
- Просят «запусти агента на репо», «почини баг автономно», «агентная задача»
- Нужен изолированный прогон: агент не должен трогать файлы вне workspace
- Нужен полный лог сессии (JSONL) для аудита каждого шага
- Нужно сравнить 2+ модели на одной задаче (по качеству и токенам)
- Нужен Web UI для интерактивной работы с агентами (
dsh web)
Do NOT use when:
- Задача простая и решается напрямую (мелкая правка, вопрос) — прямое редактирование быстрее, чем подъём агента
- Нет API-ключа и нет OpenAI-совместимого endpoint — dsh без них не запустится
- Нужна классическая оценка качества модели на бенчмарках (MMLU и т.п.) — это lm-evaluation-harness, а не dsh
- Нужен полный контроль каждого правки с апрувами — оставайся в основном агенте
📦 Files
SKILL.md— этот файлscripts/dsh_task.py— запуск одной агентной задачи (Python SDK, JSON-RPC)references/runbook.md— чек-листы, типовые ошибки, примеры конфигов
🧰 Usage
0. Требования (один раз)
# Python 3.10+; SDK:
pip install deepseek-harness-sdk
# API-ключ: DEEPSEEK_API_KEY в env…
export DEEPSEEK_API_KEY=sk-...
# …ИЛИ ключ DeepSeek из auth.json opencode (провайдер `deepseek`) —
# dsh_task.py подхватит его автоматически:
# ~/.local/share/opencode/auth.json → ~/.config/opencode/auth.json
# ИЛИ свой OpenAI-совместимый endpoint (vLLM и т.п.):
# export DEEPSEEK_BASE_URL=http://127.0.0.1:8000/v1
# Модель (по умолчанию deepseek-v4-flash):
# export DSH_MODEL=deepseek-v4-flash
1. Одна задача в изолированном workspace
python3 skills/dsh-runner/scripts/dsh_task.py \
--workspace /tmp/agent-ws/dj1 \
--session-root /tmp/agent-sessions \
--session-id fix-001 \
"В репозитории падает тест test_player.test.ts. Найди причину и почини."
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 9d ago First seen · 130 lines · 260 tokens per session scan A f502d3475555
dsh-runner is a skill published in the GitHub repository bestdeejay-design/agent-skills (5 stars, last pushed yesterday), licensed MIT. It adds 260 tokens to every session and 1,882 once invoked, about $0.0013 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 2026-08-31.
Other skills, from other repositories
workflow
Use when a task is too large for turn-by-turn orchestration and should run through the big-task workflow lane: system-wide changes, large migrations, repo-wide audits, high-confidence verification, or tasks explicitly asking to run a workflow. Claude Code uses native dynamic workflows; Codex, OpenCode, and Grok use…
skill-compiler
Automatic solved-to-skill compiler — detects novel task completions and autonomously drafts new SKILL.md files. Stolen from Hermes Agent's learning loop (NousResearch, 2026-05-11).
context-compactor
9-section context compression with analysis scratchpad. Adapted from Claude Code's /compact system (2026-03-31).
daemon-loop
Autonomous recurring agent tasks — converts workflows into persistent background daemons that run on intervals. Stolen from Boris Cherny's Claude Code /loop pattern (2026-03-31).
trade-journal-analyzer
Unified post-trade analytics: journal pattern extraction + drawdown classification. Absorbs: drawdown-classifier.
Deep Research Loop
Multi-step web research, compilation, and synthesis workflow. Scrapes multiple sources, cross-references claims, and produces a structured research brief.