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 myfmarco-arch/ai-research-detective --skill research-archivistgit clone --depth 1 https://github.com/myfmarco-arch/ai-research-detectiveWrote 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/myfmarco-arch/ai-research-detective/research-archivist)<a href="https://agentmods.dev/skills/myfmarco-arch/ai-research-detective/research-archivist"><img src="https://agentmods.dev/badge/skills/myfmarco-arch/ai-research-detective/research-archivist/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/myfmarco-arch/ai-research-detective/research-archivist"><img src="https://agentmods.dev/badge/skills/myfmarco-arch/ai-research-detective/research-archivist.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 | $0.00036 | $0.02164 |
| Opus 5 | $0.00018 | $0.01082 |
| Sonnet 5 | $0.00007 | $0.00433 |
| Haiku 4.5 | $0.00004 | $0.00216 |
Grade A, and why
research-archivist 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 12d 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 — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
研究知识入库助手（Archivist）
你是研究资料的入库处理器。你的工作是将原始研究资料（访谈、问卷、反馈等）逐份阅读、理解、提取关键信息，整合进一个持久化的 wiki 知识库。
方法论出处：本 skill 采用 LLM_wiki 方法论（基于 Andrej Karpathy 的 llm-wiki gist 在研究分析场景的特化）——把原始资料编译成 LLM 友好的结构化知识库，让后续分析直接在"编译好的知识"上工作，而不是每次重读原文；wiki 随每次分析/审查持续生长。
这个 wiki 是为 research-detective 侦探分析 skill 准备的——侦探在 wiki 上工作，不需要回到原始资料。你的入库质量直接决定侦探分析的质量。
直接调用参数
如果用户用 /research-archivist $source_path 调用,先把 $source_path 当作本次待入库材料或目录候选。必须先验证路径存在并说明将处理的范围;路径不存在或含义不清时停下来问。即使提供了路径,也不能跳过步骤 1 环境门禁。
核心原则
- LLM 阅读，不是 Python 关键词匹配。每份资料必须由你直接阅读理解，不能用脚本替代
- 增量处理。新资料进来时，更新已有 wiki 页面，不是重建
- 矛盾即时标记。入库时发现与已有知识矛盾的内容，立即记录到矛盾页
- 未归类的不丢弃。无法归入任何主题的观察，放进待审页——这是侦探盲区扫描的输入
- wiki 随分析生长。每次
research-detective侦探分析、research-reviewer对抗审查产生的新涌现（新主题、新矛盾、新关联、被反驳的理论、被发现的盲区），都会回写到 wiki。wiki 不是只在入库时变化，而是随着分析不断变厚——这是与传统知识库的核心区别。回写规则见 ../../contracts/analysis_writeback.md。
工作流程
步骤 1：初始化（环境门禁，不可跳过）
这是硬门禁，不是建议。 被唤起后，无论用户多急、
data/里是否已有资料，你必须先走完本步再决定下一步。严禁看到data/有内容就默认"增量更新"直接跳到步骤 3 入库——跳过门禁 = 本次入库作废。门禁的目的：进入步骤 2 之前，确保你已理解研究问题、入库边界，且 CONTEXT / README / CLAUDE.md 三件套就位。
① 探测目录状态——检查 CONTEXT.md（研究背景/问题，单一真源）、README.md（入库范围/边界/局限）、项目根 CLAUDE.md（项目级硬约束）、wiki/（已有知识库）是否存在。
② 按下表对号入座（CONTEXT × wiki 的有无覆盖全部四种状态，这是初始化分支的唯一真源）。如果 cold_start 识别到旧报告/PPT/memo/研究计划,按其材料分层处理:旧报告/PPT/memo 是二手分析或待验证假设,研究计划是项目语境,都不能当作一手资料入库。
CONTEXT.md |
wiki/ |
判定 | 动作 |
|---|---|---|---|
| 无 | 无 | 首次入库 | 走 ../../shared/cold_start.md 完整流程（扫项目 → 生成 CONTEXT/README 待确认草案 → 一次性请用户补齐并校对 → 用户确认后合并写入 → 配置 CLAUDE.md），再做下方③④。完成前不许读 data/ 做提取 |
| 无 | 有 | 异常态（wiki 在但 CONTEXT 丢了） | 停下，告诉用户"检测到 wiki 但缺 CONTEXT.md"，按 cold_start 补齐 CONTEXT/README（同样先展示草案、用户确认后再写入），再做③④ |
| 有 | 无 | 已配置未入库 | 跳过冷启动，做③④ |
| 有 | 有 | 增量更新 | 跳过冷启动，做③（读 wiki/_index.md 了解已处理资料和主题）+ ④ |
③ 完整性检查（凡 CONTEXT.md 已存在就必跑，红线阻断）：
- 读
CONTEXT.md的速读卡、我的身份、研究问题、底线——决定本次入库的视角和颗粒度（同样的访谈，研究问题不同，提取的主题颗粒度不同）；读README.md的入库范围与边界——避免范围外资料混入 - 跑
python3 ${CLAUDE_SKILL_DIR}/../../shared/scripts/lint_context.py CONTEXT.md：红线非 0（占位符残留 / 必填字段空 / 核心问题 < 20 字）→ 停下按 cold_start 让用户补齐，红线清零前不前进；仅黄线（底线套话 / 填充式动词）→ 提示改写但不阻断 - 检查项目根
CLAUDE.md：缺失或非本 skill 版本 → 按 ../../shared/cold_start.md 步骤 4 第 5 项处理（自动复制或追加，先征求用户同意）
What ships with it
17 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.
- guides/wiki_quality_rules.md 2.6 KB
- scripts/lint_source_coverage.py 3.2 KB runs code
- scripts/tests/fixture_wiki_bad/data/interview_01.md 889 B
- scripts/tests/fixture_wiki_bad/data/interview_02.md 644 B
- scripts/tests/fixture_wiki_bad/wiki/_source_coverage.md 159 B
- scripts/tests/fixture_wiki_bad/wiki/quotes.md 205 B
- scripts/tests/fixture_wiki_bad/wiki/themes/theme_push_fatigue.md 549 B
- scripts/tests/fixture_wiki_good/data/interview_01.md 889 B
- scripts/tests/fixture_wiki_good/data/interview_02.md 644 B
- scripts/tests/fixture_wiki_good/wiki/_source_coverage.md 457 B
- scripts/tests/fixture_wiki_good/wiki/quotes.md 377 B
- scripts/tests/fixture_wiki_good/wiki/themes/theme_push_fatigue.md 752 B
- scripts/tests/run_tests.sh 4.8 KB runs code
- scripts/verify_quotes.py 11 KB runs code
- workflows/incremental_update.md 1.3 KB
- workflows/intake_validation.md 3.3 KB
- workflows/intake_workflow.md 9.0 KB
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.
- 12d ago First seen · 76 lines · 36 tokens per session scan A 754360c88647
research-archivist is a skill published in the GitHub repository myfmarco-arch/ai-research-detective (2 stars, last pushed 2mo ago), licensed MIT. It adds 36 tokens to every session and 2,164 once invoked, about $0.0002 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.
Other skills, from other repositories
vault
Generate an Obsidian LLM wiki vault from a project's source code and any static documents in raw/.
rlm
This skill should be used when the user asks to "process a large file", "analyze document exceeding context", "use RLM", "recursive language model workflow", "chunk and analyze a file", "handle long context", or needs to work with documents too large to fit in the context window. Orchestrates the full RLM loop using…
dcf-model
Build discounted cash flow valuation workbooks in Excel.
structuring-radiology-reports
Converts free-text radiology narratives into structured findings and impression — with measurements, laterality, anatomy, and follow-up recommendations — after OpenMed NER. Use when the user has a CT/MRI/X-ray/ultrasound/mammography report and needs the sections split (technique, comparison, findings, impression)…
skill-doc-delivery
Convert markdown to DOCX, PPTX, XLSX, PDF office documents — use when you need exportable deliverables.
parse-document
Convert a PDF, scan, image of a page, or office file to clean markdown through the connected Superlinked MCP edge, so the source document is not read into model context directly. Use when the user asks to read, parse, OCR, extract from, summarize, or answer questions about a document.