deep-research
Deep research harness — fan-out web searches, fetch sources, adversarially verify claims, synthesize a cited report.
- License 许可
- MIT
- 最近更新
- 2026-08-24
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- deep-research
- description
- Deep research harness — fan-out web searches, fetch sources, adversarially verify claims, synthesize a cited report.
- when_to_use
- When the user wants a deep, multi-source, fact-checked research report on any topic. BEFORE invoking, check if the question is specific enough to research directly — if underspecified (e.g., "what car to buy" without budget/use-case/region), ask 2-3 clarifying questions to narrow scope. Then pass the refined question as args, weaving the answers in.
02 资源文件 1
- 脚本scripts/workflow-script.js17.3 KB
03 SKILL.md 正文
Run the "deep-research" workflow.
Deep research harness — fan-out web searches, fetch sources, adversarially verify claims, synthesize a cited report.
When the user wants a deep, multi-source, fact-checked research report on any topic. BEFORE invoking, check if the question is specific enough to research directly — if underspecified (e.g., "what car to buy" without budget/use-case/region), ask 2-3 clarifying questions to narrow scope. Then pass the refined question as args, weaving the answers in.
Phases:
- Scope: Decompose question (from args) into 5 search angles
- Search: 5 parallel WebSearch agents, one per angle
- Fetch: URL-dedup, fetch top 15 sources, extract falsifiable claims
- Verify: 3-vote adversarial verification per claim (need 2/3 refutes to kill)
- Synthesize: Merge semantic dupes, rank by confidence, cite sources
Invoke: Workflow({ name: "deep-research" })
Workflow Script
04 同仓库更多 Skills
{{SKILL_ID}}
{{ONE_SENTENCE_ACTIVATION_DESCRIPTION}}
session-handoff
Generate rigorous, multi-artifact session handoff packages that preserve full context for zero-loss context switching between AI sessions. Use this skill whenever the user asks to: save session state, create a handoff, package context for a new session, summarize and preserve a conversation for later, do a session dump, create a context transfer, prepare a state snapshot, export session knowledge, hand off to another agent, persist session memory, "wrap up" or "save where we are", create a knowledge package, synthesize session context, or produce a resumption artifact. Also trigger when the user says things like "save this for next time", "capture everything we discussed", "I need to pick this up later", "package this session", "make sure nothing is lost", "create a handoff document", "session export", "context preservation", or any variant of persisting conversational and working state. Even if the user just says "handoff" or "save state" — use this skill.
vdm-publication-bundle
Generate, audit, and validate complete VDM research bundles following theorem-bearing manuscript standards. Use this skill when the user wants to: create a research bundle, Lean4 package from formal claims, SymPy adversarial symbolic audit script, a skeptical-reviewer Jupyter notebook, score papers w/ CRAFT/BASS review metrics, audit burden coverage, generate SHA256 manifests, create figures following standards, scaffold a new Complete Formalism publication from claims, or check whether existing bundles meet certificate requirements. Also trigger when the user mentions "publication bundle","validation bundle","CRAFT score","BASS score","burden coverage","claim attack","gate check","reviewer notebook","adversarial audit","publication certificate","closure certificate","artifact manifest", "figure standards", "review-ready", or any CF/CFN publication workflow. Even if the user just says "score this paper"/"check if this is publication-ready", use this skill.
Use when tasks involve reading, creating, or reviewing PDF files where rendering and layout matter; prefer visual checks by rendering pages (Poppler) and use Python tools such as `reportlab`, `pdfplumber`, and `pypdf` for generation and extraction.
Use when tasks involve reading, creating, or reviewing PDF files where rendering and layout matter; prefer visual checks by rendering pages (Poppler) and use Python tools such as `reportlab`, `pdfplumber`, and `pypdf` for generation and extraction.
template-skill
Replace with description of the skill and when Claude should use it.