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 XRenSiu/claude-code-forge --skill skill-evolvegit clone --depth 1 https://github.com/XRenSiu/claude-code-forgeWrote 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/xrensiu/claude-code-forge/skill-evolve)<a href="https://agentmods.dev/skills/xrensiu/claude-code-forge/skill-evolve"><img src="https://agentmods.dev/badge/skills/xrensiu/claude-code-forge/skill-evolve/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/xrensiu/claude-code-forge/skill-evolve"><img src="https://agentmods.dev/badge/skills/xrensiu/claude-code-forge/skill-evolve.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 | $0.00240 | $0.04250 |
| Opus 5.5 | $0.00096 | $0.01700 |
| Sonnet 5.5 | $0.00048 | $0.00850 |
| Haiku 4.5 | $0.00024 | $0.00425 |
Grade A, and why
skill-evolve 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 1mo 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 — 219 lines — stays where its author put it; the contents beside it link to each section on GitHub.
skill-evolve
让一个 skill 从 1 分变 100 分,不靠灵感,靠循环。
Announce at start: "I'm using the skill-evolve skill to run the 9-dimension rubric (with no-skill baseline comparison) + ratchet hill-climbing loop on the target SKILL.md."
Skill(结构化 Prompt)和 ML 超参数一样,是可以被自动化搜索优化的参数空间。
Skills are searchable parameter spaces — same as ML hyperparameters.
它解决什么问题
写完一个 SKILL.md 之后,绝大多数人停在「能跑就行」。但「能跑」和「好用」之间有一条 30 分的鸿沟:触发不准、步骤含糊、边界没定义、激活了不知道先干啥。
人工反复打磨成本极高,且改一处不知道是不是变差了。本 skill 把这个过程自动化 + 防退化:
- 自动化:9 维 rubric 直接告诉你哪里弱、改哪里
- 防退化:每次改完独立打分,不升反降立刻回滚,分数只能涨不能跌(棘轮)
- 防自欺:rubric 的三个高信号维(失败机制 / 可执行性 / 高风险黑名单)来自 SkillLens 实测,且每轮强制跑「带 skill vs 不带 skill」基线对比——刷分但实际拖累 agent(负迁移)会被当场抓出并回滚
灵感来自 Karpathy autoresearch(修改→训练→保留/回滚)、alchaincyf/darwin-skill(把范式迁移到 Skill 优化),以及微软 SkillLens(arXiv 2605.23899,46.4%→73.8% 的三维药方 + 负迁移发现)/ SkillOpt(arXiv 2605.23904,拒绝缓冲 + 慢更新记忆 + 文本学习率三个稳定性控制)。
输入与输出
输入:
- 目标 skill 路径(必须包含
SKILL.md),例如plugins/foo/skills/bar/ - (可选)3 个已知测试 prompt + 期望输出方向(无则自动从 SKILL.md 推断)
- (可选)目标分数(默认 90)/ 最大轮次(默认 15)/ 收敛阈值(默认连续 3 轮无提升)
输出:
.skill-evolve/<skill>/experiments.tsv—— 每轮:{round, dim_scores(9), total, negative_transfer, change_summary, kept_or_reverted, flags(dry_run/single_judge), commit_hash}.skill-evolve/<skill>/baseline.md—— 初评全维度分数 + 最弱维度诊断 + 确认过的测试集.skill-evolve/<skill>/learnings.md/dead-ends.md—— 滚动正/负反馈记忆.skill-evolve/<skill>/final-report.md—— 终评 + 改进清单 + 前后对比 + 负迁移确认- 每轮
experiment: <desc>前缀的 git commit(棘轮历史)
触发后立即做什么
1. 确认目标 skill 路径存在且包含 SKILL.md
2. 检查 git working tree 是否干净(脏树先暂存或提示用户处理)
3. 创建工作目录 .skill-evolve/<skill-name>/ 存放日志和报告
4. 初始化滚动记忆(移植自 SkillOpt):
: > .skill-evolve/<skill-name>/learnings.md # 已验证规律(正反馈)
: > .skill-evolve/<skill-name>/dead-ends.md # 已证伪改法(负反馈,禁重试)
5. 检查 Agent/subagent 工具是否可用;不可用 → 走 dry_run 降级模式(见 test-protocol)
6. 进入 Phase 0
Phase 0.5:测试 prompt 设计与确认(先于评估的门)
测试 prompt 的质量直接决定优化方向对不对(SkillLens:测试太弱会导致「分数虚高、实战还烂」)。所以在任何评估之前:
- 按 test-protocol「测试 prompt 自动生成」规则备出 3 个测试 prompt(happy path)+ 1 个边缘 prompt(相邻领域、超范围)
- 写入
baseline.md的「测试集」区,展示给用户确认 - 用户可替换/补充更刁钻的 case。用户确认后才进入 Phase 0——这一步不许跳过
What ships with it
7 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.
- 1mo ago First seen · 219 lines · 240 tokens per session scan A a7e17a413f6a
skill-evolve is a skill published in the GitHub repository XRenSiu/claude-code-forge (2 stars, last pushed 29d ago), licensed MIT. It adds 240 tokens to every session and 4,250 once invoked, about $0.0010 per session on Opus 5.5. 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.
Other skills, from other repositories
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
tika-eval-compare
Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".
cua-driver
Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.
atmos-validation
Validate Atmos projects, components, arbitrary JSON Schema inputs, EditorConfig, and GitHub Actions; use affected-file selection and native CI annotations.
skill-benchmark
Benchmark AI skill effectiveness by measuring implementation quality against legacy constraints.