eval-session

A review process for a completed sid-code coding session. It examines the result, the language model's steps, and the test harness, then records evidence-backed bugs and improvement opportunities.

In plain words
What is it for?
Investigating a completed session, finding bugs or wasted effort, documenting evidence with file and line references, requesting human confirmation, and creating traceable follow-up fixes.
Why use it?
It helps separate problems in sid-code from mistakes or inefficiencies in one task run. A human reviews the findings before they become a fix list, so unsupported changes are not treated as requirements.

Skill for Claude CodeCodex

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.

agentmods
npx agentmods add skills/rushengzhou/sid-code/eval-session
Any agent
npx skills add rushengzhou/sid-code --skill eval-session
Clone the repo
git clone --depth 1 https://github.com/rushengzhou/sid-code

Made for: Claude Code, Codex.

Per session 245 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 10,911 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00245 $0.10911
Opus 5 $0.00122 $0.05456
Sonnet 5 $0.00049 $0.02182
Haiku 4.5 $0.00024 $0.01091

Measured yesterday against content hash eaadd6ae1096, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

eval-session 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 yesterday.

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.

.claude/skills/eval-session/SKILL.md · 306 lines

How it starts

The opening of the file, as written. The whole thing — 306 lines — stays where its author put it; the contents beside it link to each section on GitHub.

eval-session:sid-code 任务会话三段式评估

目的锚点(每次执行前先默念一遍,防止跑偏)

通过一次会话的轨迹,做结果评估 + 过程评估(LLM + harness),找出 sid-code 的 bug 和可优化地方,驱动后续优化。

这句话里有两个常被做丢的重点:

  1. 是"bug 可优化地方",不是只找 bug。 任务跑对了但绕了远路、花了冤枉钱、某个 nudge 太弱、缺一个能力——这些都不是缺陷,却正是 sid-code 该优化的地方。只盯"崩没崩"会系统性漏掉一半价值。
  2. 是"找出 sid-code 的问题",不是"评模型考了几分"。 LLM 过程评估不是目的,是手段——模型的每一次失误/低效,都要回问一句"harness 本可以拦住/让它更容易做对吗?",能翻译成 harness 改进的才是产出。评模型本身,评完就作废了。

术语统一:本 skill 把最终产出统称发现(finding),分两类——缺陷(bug)(行为错误,该修)与优化点(opportunity)(不算错但值得改进:效率/成本/引导强度/能力缺失)。§3 报告两类都要出,别把优化点硬塞进"缺陷"栏,也别丢掉。

核心原则:结果先行,fix 是派生物

评估结果本身就是交付物,不是通往 todo-fix 的过场。 顺序不可颠倒:

三段评估结果(主交付物,要写透)
   ↓  人类复核 + 二次确认(gate:未确认不进入下一步)
todo-fix 清单(派生物,每条都必须回指某条已确认的评估结论)
   ↓  修复(另开会话,本 skill 不做)
验证修改生效(按 todo-fix 里预置的"验证方法"逐条确认)

三条硬约束:

  1. 评估结果要写透,不能压成一行表格。 每段用叙述 + 证据展开,像给人看的分析报告,而不是待办清单的附录。宁可结果段长,也不要为了赶 todo-fix 而把判断依据省略。正向样本、边界、"措辞超出证据"这类细节都要保留——它们正是结果可信度的来源。
  2. todo-fix 的每一条都要可追溯到评估结论。 格式:fix 项 ← 源自 §X 发现 N。没有对应评估结论的 fix 项不许出现(否则就是凭空臆想的改动)。
  3. 人类是决策者,skill 只提供依据。 报告必须专门列出"需要人类二次确认的判断项"(见 Phase 6),这些点没拿到人类确认之前,不能当成定论驱动修复。

工具与大模型的职责分层(谁产事实、谁下结论)

评估既依赖脚本也依赖大模型,但分工必须清晰——混了就会要么幻觉数字,要么把工具误报当结论:

谁做 产出 铁规则
确定性层 trace-digest + 手写 bash/python 计数、解析、配对、聚合等可复现的事实 + 假设 大模型不许口算/臆造数字;该由工具算的必须真跑
验证层 大模型 把工具的假设逐条拿去 read 原始数据 + read 源码,证实/证伪 工具的假设未经此层验证不许进报告;拒掉假阳性
判断层 大模型 严重度、"缺陷 vs 取舍 vs 优化点"边界、修复方向、MEMORY 关联 任何脚本都做不了,只能大模型下

一句话:工具产"事实+假设",大模型产"结论"。任何结论都不能停在工具未验证的输出上;任何该由工具算的数字都不能由大模型口算。 脚本出错或遗漏时,由验证层兜住——而且脚本自身的缺陷就是一条 harness 发现(如"并行子代理误报循环"),按 §3 写进报告,流程因此自我改进。

目的与边界

对一次已完成的 sid-code 任务会话做系统评估,核心目的是找出 sid-code(harness/工具本身)的 bug 和可优化地方

本 skill 只评估、只出报告,自己不改任何代码。 针对 sid-code 缺陷/优化点的实施在另一个会话做——本 skill 要为它铺好两件事:① 每个发现预置"验证方法"(修完怎么确认生效);② 明确哪些结论需人类拍板。

区分两个"改代码": "本 skill 不改代码"指的是评估动作本身不动 sid-code 源码。但被评估的那次会话可能已经做了代码改动(如任务含"修复所有缺口")——那些改动是交付物,恰恰是结果评估要严查的对象(见 Phase 2 代码改动五关)。别因为"不改代码"就跳过对被评会话已落地修复的核查。

Read the full file on GitHub · 306 lines

Files

What ships with it

2 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.

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. yesterday First seen · 306 lines · 245 tokens per session scan A eaadd6ae1096

Subscribe to this mod's changes

eval-session is a skill published in the GitHub repository rushengzhou/sid-code (2 stars, last pushed 2d ago), licensed MIT. It adds 245 tokens to every session and 10,911 once invoked, about $0.0012 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.

Related

Other skills, from other repositories

omd-contract

把审议结论结晶成正式契约文档落盘 docs/plan/,当 /omd-execute 的执行契约。承接 /omd-grill 的决策记录表,写给没有对话上下文的执行器看。含 crystallize/crystals 职能。Trigger:/omd-contract、定契约、写成执行契约、写成 SDD、SDD、结晶、方案定了记下来、列结晶。.

AbyssCN/oh-my-dag · 105 tokens

omd-grill

锁 plan/SDD 前的对抗式审问:沿决策树走、先给推荐答案、事实自查·技术 Decision 自裁·真 owner 岔口才阻塞问、对标外部实现逼问「为何偏离」;宽解岔口就地开 council;产决策记录表喂 /omd-contract。审议纪律:只讨论不动手。Trigger:/omd-grill、审问、盘问这个方案、把这事讨论清楚、压测计划、stress-test。.

AbyssCN/oh-my-dag · 122 tokens

omd-video

视频→逐段结构化笔记 (MiMo-v2.5 原生吃画面+音频, 非 whisper 转写; 可重入管线)。讲解/课程视频里 PPT 框架图/代码/提示词是画面独有、音频拿不到的信息。产 ALL-NOTES.md 交 /omd-council 或 dagresearch 做综合。Trigger:/omd-video、抖音/B站/YouTube 讲解视频、课程系列、把这些视频学一遍/提炼、画面里有代码/图表/PPT。Skip:文字原文综合→/omd-council;网页内容→dagresearch(检索版)。.

AbyssCN/oh-my-dag · 155 tokens

omd-docs-drift

文档漂移追踪:确定性死路径/死锚闸(bun test 常驻)之外的语义半——按 docs/docs-map.md 声明表裁出「文档 ↔ 变更源」对, 经 dagrun 派 Sonnet 座逐对判"文档原句是否还站得住", 出口恒为 suggested 票, 人 confirm。Trigger:/omd-docs-drift、查文档漂移、文档跟没跟上代码、docs drift。.

AbyssCN/oh-my-dag · 119 tokens

omd-audit

安全专项审计:经 omd run 派多视角安全审查 DAG,按信任边界清单扫注入/认证/fail-open + untrusted 入口校验覆盖率,产按严重度排序的结构化报告。Trigger:/omd-audit、安全审计、查漏洞、信任边界、security review。.

AbyssCN/oh-my-dag · 80 tokens

omd-research-deep

终极档深度调研:种子作者化多角度抓取 + council 分解 + 多轮缺口补挖,一条工具调用出整领域 grounded 报告。Trigger:/omd-research-deep、深度调研、deep research、整领域调研、调研到 grounded 底座、把这个领域研究透。Skip:轻量单点查证(tavily 直查)/ 单问题综合(dagresearch 默认档)/ 代码理解(dag-map)。.

AbyssCN/oh-my-dag · 117 tokens