Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Mainline

ASecurity

叙事主线核查(测量原件,只产 finding,不改稿):建 paper-level purpose record 与 contribution graph,再按 cold reader 回答八问,定位读者需要回溯、补隐藏上下文或在竞争解读间抉择的地方。findings 走 sci-paper.feedback.v1;不假设三幕模板,多贡献论文可有多个明确相关的分支,但都从属于唯一的 thesis line。可单独运行,也是 /sci-paper:paper-review 维度 E 的唯一来源。Use for: 叙事主线, 主线检查, contribution graph, 冷读八问, narrative spine, 文章结构是否讲得通, 读者会不会卡住。

5 stars
0 votes
0 copies
0 views
Added 9/28/2026
ai-agentsnode

Security Analysis

A100/100

Scanned 9/28/2026

Install to Claude Code

$npx -y skills add skymanbp/sci-paper --skill mainline --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Mainline?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Mainline
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/skymanbp-mainline/badge)](https://www.skillsdirectory.com/skills/skymanbp-mainline)

More formats (shields.io, HTML) on the badges page.

Files
SKILL.md
---
name: mainline
description: 叙事主线核查(测量原件,只产 finding,不改稿):建 paper-level purpose record 与 contribution graph,再按 cold reader 回答八问,定位读者需要回溯、补隐藏上下文或在竞争解读间抉择的地方。findings 走 sci-paper.feedback.v1;不假设三幕模板,多贡献论文可有多个明确相关的分支,但都从属于唯一的 thesis line。可单独运行,也是 /sci-paper:paper-review 维度 E 的唯一来源。Use for: 叙事主线, 主线检查, contribution graph, 冷读八问, narrative spine, 文章结构是否讲得通, 读者会不会卡住。
disable-model-invocation: false
argument-hint: "<file_path> [--field <name>]"
---

# mainline — 叙事主线核查(测量原件)

> **Normative authority:** `docs/SCIPAPER_STANDARD.md` §5.4(thesis spine)。
> findings 一律用 `sci-paper.feedback.v1` 的后果类别与 measurement state。
> `/sci-paper:paper` 定义主线该长成什么样;本 skill 是它的**审查端**。

**这是测量原件:只产 finding,不改稿。** 结构修复的动作由组合 skill 路由到
`/sci-paper:de-ai` Pass 3(改写)或 `/sci-paper:condense`(删减合并)。

## 0. 取证纪律

1. **必须冷读。** 从头到尾 Read 全文;不得继承任何"这段我懂了"的既有结论。
   本 skill 的全部价值来自模拟一个第一次读这篇论文的人。
2. **每个节点带 `file:line`。** purpose record 与 contribution graph 的每个条目
   都必须指回当轮读到的具体位置,不能是概括。
3. **困惑必须可复现。** 写下"读者会卡住"时,要能指出卡在哪一句、缺的是哪个前提。
   说不出来的困惑不是 finding。
4. **拿不到就标 unmeasured。** 无法访问 companion paper、数据或图时显式说明,
   不能因为读不到就当作没问题。

## 1. 第一步:purpose record

建 paper-level purpose record,全部带 `file:line`:

- **root question** —— 这篇论文到底在问什么?
- **contributions[]** —— 有哪几项贡献?
- 每项贡献的 **method_or_argument** —— 靠什么方法或论证成立?
- 每项贡献的 **key_evidence[]** —— 关键证据是什么?
- **take-home + scope** —— 读者应该带走什么,以及这个结论覆盖到哪为止?

## 2. 第二步:contribution graph

- **nodes** = claims / definitions / methods / evidence / conclusions
- **有向边** = prerequisite / supports / qualifies / contrasts / derives-from

判读规则:

- **不预设三幕。** motivation → method → validation 是常见默认,不是所有论文
  必须套用的模板。
- **多贡献合法。** 多贡献论文可以有多个明确相关的分支;问题是**关系有没有讲
  清楚**,而不是图是否恰好只有一个连通分量。但每个分支都必须从属于唯一的 thesis
  line(标准 §5.4:全文只有一个中心结果,可用一句话说出)。
- **disconnected node 是待查证据,不是自动 fatal。**
- **禁手**:用连接词桥接逻辑断链(拿 Furthermore / Moreover 顶替缺失的前提)。

## 3. 第三步:cold reader 八问

按第一次读的读者依次回答:

1. root question 是什么?
2. 贡献有哪些?
3. 贡献之间是什么关系?
4. 各自的方法 / 论证是什么?
5. 各自的关键证据是什么?
6. take-home 与 scope 是什么?
7. thesis line 是什么?能否用一句话说出唯一的中心结果,且每个贡献分支都从属于它?
8. 读者在哪里需要**回溯**、**补隐藏上下文**,或在**竞争解读**之间抉择?

前七问答不上来,就是主线本身的缺陷,而不是读者的问题。第八问的每个落点都是
一条候选 finding。

## 4. 后果分类

| 情形 | 类别 |
|---|---|
| 矛盾,或关键 claim 缺必要支撑论证 | `integrity_blocker` |
| 高曝光、可复现的困惑 + 能给出具体修复 | `advisory`,排在最前 |
| 局部措辞、可读性偏好 | `advisory` |

- 缺失支撑关键 claim 的必要论证可升为 `integrity_blocker`。
- 其余 narrative finding 都是普通 advisory,按证据强度和读者曝光排序(标准 §4)。
  不升 strong:strong 要求效应超出校准过的操作点,冷读没有操作点。
- **不得要求所有结构 advisory 归零。** 普通 residual 报告即可;作者可以逐条处置,
  但这不是收尾条件。

## 5. 边界 —— 本 skill 不做什么

- **不做冗余判定。** 重复 claim、无信息增量段、死定义归 `/sci-paper:condense`
  的检测端与 `/sci-paper:paper-review` 维度 I,不在此复述。
- **不做逻辑/统计审查。** claim graph 里的循环论证、偷换条件、统计前提归
  `/sci-paper:logic`。本 skill 只问"读者能不能跟上",不问"论证是否成立"。
- **不做语言/语域审查。** 归 `/sci-paper:de-ai`。
- **不做跨章节一致性核对。** 同一符号/常数/样本量跨章节是否一致归
  `/sci-paper:paper-review` 维度 L。
- **不改稿。** 见顶部。

## 6. 报告

```markdown
# Mainline — Narrative Spine Findings

**Target**: <file> | **Field**: <field or none>

## Purpose record
| field | value | file:line |

## Contribution graph
| node | type | file:line | edges (type → target) |

## Cold-read eight questions
| # | question | answer or gap | file:line |

## Findings
| id | kind | location | reader confusion | proposed fix | disposition |
```

不打印 PASS/FAIL 行。终止态是八问都有明确答案或明确记录的缺口。

## 7. 反模式

- ❌ "整体看结构还行" —— 没有 purpose record 和八问逐条答案的总评无效。
- ❌ 拿摘要或 intro 概括代替全文冷读。
- ❌ 因为图不是单连通就判 fatal。
- ❌ 要求每篇论文都是 motivation → method → validation。
- ❌ 把"我读着有点乱"写成 finding,却指不出卡在哪一句。
- ❌ 复述 `/sci-paper:logic` 或 `/sci-paper:condense` 已经拥有的检查项。

Attribution

skymanbpskymanbp
View sourceMore from skymanbp →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Ultra-compressed communication mode that cuts output tokens while keeping technical accuracy. Levels: lite, full, ultra and the wenyan variants. Use for /caveman, "caveman mode", "talk like caveman", "be brief" or "less tokens".

1074701 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

695601 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3351 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

691 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →