菜单
Normalize raw deep research outputs into a compact Open Knowledge Format bundle with source, finding, uncertainty, method, index, and log concepts. Use after provider runs when reports need portable structured markdown and citation-preserving validation.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Prepare and run bounded calls across deep research provider APIs with explicit cost limits, API-key sources, benchmark prompts, and output capture. Use when comparing providers, creating reproducible API commands, or normalizing raw provider outputs after a research run.
Run Exa-backed deep research, neural search, and page-content retrieval with bounded result and page-read counts. Use when a task needs Exa Search, Exa Contents, or Exa Research inside a source-discovery or benchmark workflow.
Run Google Gemini deep research and search-grounded Gemini research tasks with bounded scope and captured outputs. Use when research should rely on Gemini, Google Search grounding, or Google-hosted document context.
Run GPT Researcher as a self-hosted open-source deep research framework using an LLM plus a search provider. Use when a task needs inspectable local report generation, configurable retrievers, or comparison against commercial research APIs.
Use Jina Reader and Jina Search APIs for deep research retrieval, URL-to-markdown conversion, and markdown-oriented search results. Use when a workflow needs clean page extraction, search snippets, or reader output before synthesis.
Run LangChain Open Deep Research or LangGraph-based iterative research workflows with configurable retrieval, reflection, and server-backed execution. Use when a benchmark or task needs inspectable graph orchestration rather than a single hosted research API.
| name | deep-research-okf-normalize |
| description | Normalize raw deep research outputs into a compact Open Knowledge Format bundle with source, finding, uncertainty, method, index, and log concepts. Use after provider runs when reports need portable structured markdown and citation-preserving validation. |
Use this skill after a deep research agent returns a report. It turns a raw
report into a portable markdown bundle aligned with Google's Open Knowledge
Format draft: markdown files with YAML frontmatter, a required type field,
optional index.md and log.md, and citations preserved as links where
possible.
python .agents/skills/deep-research-okf-normalize/scripts/normalize_to_okf.py \
--input benchmark/results/perplexity-sonar/meta/q1_output.txt \
--bundle-dir benchmark/results/okf/perplexity-sonar-meta
For ad-hoc text:
python .agents/skills/deep-research-okf-normalize/scripts/normalize_to_okf.py \
--title "Deepresearch the deepresearchers" \
--text "Paste or pipe the report text here" \
--bundle-dir outputs/okf/deepresearchers
The generated bundle contains:
index.md - progressive-disclosure entrypoint.report.md - the normalized source report concept.findings.md - extracted claim bullets and citation-bearing lines.uncertainties.md - caveats, conflicts, missing evidence, and stale-data
risks.method.md - provider, prompt, timestamp, and normalization notes.log.md - update history.python .agents/skills/deep-research-okf-normalize/scripts/validate_okf.py \
benchmark/results/okf/perplexity-sonar-meta