菜单
Benchmark a coding model on DeepSWE through Pier and OpenRouter. Use when users ask to run DeepSWE, score a model, or verify coding-agent benchmark results.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/hoanghd218/ZaloCRM --skill ak-deep-swe命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Search library/framework documentation via llms.txt (context7.com). Use for API docs, GitHub repository analysis, technical documentation lookup, latest library features.
Interview-driven advisory skill. Analyzes a prompt or URL (GitHub issue, spec, doc), scouts the codebase, interviews the user one question at a time to reframe the problem into exact requirements and goals, then delivers honest advice: what to do, what to avoid, better alternatives, benefits and trade-offs. Use when the user asks for advice, a second opinion, a sanity check, requirement reframing before committing to a direction, or wants an existing plan or design pressure-tested (grilled) one question at a time.
Browser and desktop automation through the agent-browser CLI. Use for long autonomous browsing, compact page snapshots, screenshots, form filling, login/test flows without real Chrome profile state, scraping, exploratory QA, cloud browsers, Electron apps, Slack automation, and bug hunts.
正在显示 SKILL.md
| name | ak:deep-swe |
| description | Benchmark a coding model on DeepSWE through Pier and OpenRouter. Use when users ask to run DeepSWE, score a model, or verify coding-agent benchmark results. |
| user-invocable | true |
| when_to_use | Invoke for a costed external coding-agent evaluation, not repository-local optimization. |
| category | dev-tools |
| keywords | ["benchmark","deepswe","pier","openrouter","evaluation"] |
| license | MIT |
| argument-hint | <OpenRouter model slug> |
| metadata | {"author":"agentkit","version":"1.0.0","upstream":"Pinned MIT source archive: run-deep-swe@ce70edaa26247b84c2b9491a0cdb4964f65cf3a5"} |
Run an independent coding-agent evaluation on the DeepSWE benchmark through Pier. This skill handles benchmark setup and reporting. It does not evaluate a repository metric or authorize model spend without the user's confirmation.
git clone https://github.com/datacurve-ai/deep-swe
uv tool install datacurve-pier
pier --help
pier run --help
Run the commands below from the directory that contains the cloned deep-swe folder. If you change into deep-swe instead, use tasks or tasks/ as the -p value.
DeepSWE currently documents 113 tasks and Pier as the runner. Confirm current flags and task paths with help before running because this toolchain changes: https://github.com/datacurve-ai/deep-swe https://pypi.org/project/datacurve-pier/ https://deepswe.datacurve.ai/
Verify the exact OpenRouter model slug at openrouter.ai/models.
Run a single task first. The current native model form is:
pier run -p deep-swe/tasks/<task-id> --agent mini-swe-agent \
--model openrouter/<vendor/model>
For a deterministic subset, use the installed help to confirm the task-count and sample-seed flags, then run a small fixed sample:
pier run -p deep-swe/tasks --agent mini-swe-agent \
--model openrouter/<vendor/model> --n-tasks 10 --sample-seed 0
Use a separate model-class route only when the installed Pier help explicitly documents it; do not guess a version-specific flag.
Before a full corpus run, present the exact command, expected cost exposure, and stop condition. Continue only after explicit confirmation.
Inspect the generated job directory with the current Pier commands. When the installed version provides them, use pier view, pier analyze, and pier critique to inspect the result. Report the exact command, Pier version, task count, model slug, score or reward, cost when available, and blockers. Do not submit results or contact a leaderboard without the user's request.