Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/KtKID/x-dev-pipelinenpx agentmods add skills/ktkid/x-dev-pipeline/pipeline-eval-reportWrote 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/ktkid/x-dev-pipeline/pipeline-eval-report)<a href="https://agentmods.dev/skills/ktkid/x-dev-pipeline/pipeline-eval-report"><img src="https://agentmods.dev/badge/skills/ktkid/x-dev-pipeline/pipeline-eval-report/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/ktkid/x-dev-pipeline/pipeline-eval-report"><img src="https://agentmods.dev/badge/skills/ktkid/x-dev-pipeline/pipeline-eval-report.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 | $0.00117 | $0.02644 |
| Opus 5.5 | $0.00047 | $0.01058 |
| Sonnet 5.5 | $0.00023 | $0.00529 |
| Haiku 4.5 | $0.00012 | $0.00264 |
Grade A, and why
pipeline-eval-report 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 — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Pipeline Eval Report
把一次真实 pipeline 执行转成可复核的质量、成本和归因证据。正确性门禁先判断产物能否指导下游正确完成任务;Token、耗时和文档体积随后用于比较效率。
产物
每个评测 run 写入 pipeline-data/runs/{pipeline_run_id}/:
manifest.json:任务、版本、执行器、输入与被评产物的静态身份。events.jsonl:追加式阶段事实、评分、门禁和知识事件。telemetry.json:模型调用、Token、费用、耗时、工具和 phase 归属。grading.json:rubric 分项结果、P0/P1/P2 扣分及证据。evaluation-report.md:面向人的精简结论。
历史引导 run 使用 storage_profile: bootstrap-json-v0,只保留 manifest.json 与 events.jsonl 作为进化账本。新建评测 run 一律写齐上述五个核心文件;validator 对 bootstrap profile 执行只读兼容校验。
关键失败、扣分和优化依据写入 pipeline-data/knowledge/entries/{knowledge_id}.json。完整字段和一致性规则见 references/artifact-contract.md;报告渲染前读取 assets/evaluation-report-template.md。
最短路径
- 先判定 run 身份并创建或续用
pipeline_run_id。 - 冻结题面、rubric、skill 和候选产物的路径、版本与 SHA-256。
- 从 Mavis 或其他 provider 真源提取完整执行遥测。
- 按 rubric 逐项检查语义,记录可定位证据。
- 应用正确性门禁,定位问题最早进入 pipeline 的阶段。
- 写五类 run 产物和知识条目。
- 运行
scripts/validate_run.py,通过后交付报告路径与结论。
1. 判定 pipeline run
将以下行为登记为 pipeline run:
- 生产任务、评测任务、修复任务或回放任务触发了一个或多个 pipeline phase。
- 外部 Agent 已经生成 spec、req、代码或验证产物,当前工作负责追认、评审或补齐遥测。
- 同一用户目标由
x-spec → x-req等多个 skill 连续完成;它们属于一个 run 内的多个 phase。
身份规则:
- 一个用户任务对应一个
pipeline_run_id。 - 同一任务的反馈、修订、重评分和修复继续使用原 run,并增加
stage_attempt。 - 同题目的独立重放使用新 run,并通过
parent_run_id、baseline_run_id或 comparison 字段关联。 - Mavis Session、Codex rollout 等 provider ID 保存为
source_session_id;pipeline_run_id 负责跨执行器关联事件、报告和知识。 - 追认历史执行时,先按 source ID、任务 hash 和 workspace 搜索已有 manifest,复用匹配 run。
新 ID 使用 prun-{uuid}-{sequence}。Mavis ID 具有 32 位 UUID 十六进制主体时,可以插入标准 UUID 连字符形成稳定映射;同一 source 始终得到同一候选 ID,最终仍需检查仓库内唯一性。
在分析前写 run_started 或 run_recognized 事件。后续事实按发生顺序追加,历史事件保持原文。
2. 建立证据边界
完整读取并区分:
- 题面或用户需求:定义目标行为和范围。
- evaluator-only 答案、rubric、隐藏 Harness:定义评分与反例,只供评审阶段读取。
- 候选产物:spec、req、代码、测试、verify 或 QA 证据。
- 运行事实:Session、工具调用、命令退出码、测试结果、Token 与耗时。
- pipeline 版本:skill、模板、validator、模型和仓库 SHA。
执行器与 grader 保持信息隔离。报告记录 rubric 是否对执行器隐藏。可变或仓库外的文本产物复制到 run 的 artifacts/ 后再评分;大型产物记录不可变存储引用和 SHA-256。
What ships with it
5 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 · 190 lines · 117 tokens per session scan A 8aa9f1dcc245
pipeline-eval-report is a skill published in the GitHub repository KtKID/x-dev-pipeline (12 stars, last pushed 4d ago), licensed MIT. It adds 117 tokens to every session and 2,644 once invoked, about $0.0005 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 2026-09-09.
Other skills, from other repositories
procoder
Work like a senior developer in a repository governed by procoder: run the commit gate before calling anything done, format and lint through the binary, and drive the spec, plan, todo, backlog, and sprint chain in .procoder/. Use this skill when the repository contains a .procoder/ directory or an AGENTS.md naming…
explain
Guided code tour of a file or subsystem this session touched — entry point, the load-bearing pieces, the edges, and what connects to it.
test-audit
Audit test suites for T1-T4 violations using AST analysis, mock detection, and multi-stage synthesis. Invoke when user asks to audit tests, check test quality, find mock violations, review test effectiveness, or inspect test suites for over-mocking. Triggers automatic rewrites when quality gates fail.
code-review
Comprehensive code review with distinct aspect based sections. Use when reviewing code, checking for security issues, finding type safety problems, auditing code quality, or when user asks to review code, PRs or changes. Three-phase workflow runs static tools, LLM judgment, and writes diagnostic log.
spec-drift-check
Audits a WP brief for drift against current code — extracts claims, verifies each, emits PROCEED/STOP verdict. Use when starting a new WP, before consuming a spec as binding, or when a doc references paths/lines/functions.
next-steps
Have the Product Owner survey the project's current state — shipped features, in-flight work, open findings, standing goals — and propose one concrete next feature ready to hand to /story-time or /spark. Use when the user has no idea in hand and asks "what should we build next", when a loop just closed and the next…