From 10x-swe
Use the Readwise CLI to run relevance-first research across Reader documents and highlights, then synthesize an Obsidian-ready note with source-linked evidence and explicit confidence gaps. Use when researching topics from Readwise content for PKM workflows in Obsidian.
npx claudepluginhub cercova-studios/terminal-agent-plugins --plugin 10x-sweThis skill uses the workspace's default tool permissions.
<objective>
Queries Readwise for highlights, quotes, annotations, full document text, and article content. Adds highlights or tagged documents to notebooks. Auto-activates on search or fetch requests.
Accesses Readwise highlights and Reader documents via CLI for searching, listing, reading, and creating highlights from the command line. Useful for terminal integration with reading libraries.
Ingests source material from Obsidian vaults, markdown notes, documents, and URLs; builds knowledge maps and identifies research gaps for writing projects.
Share bugs, ideas, or general feedback.
Scope for this skill is intentionally narrow: search, retrieval, evidence ranking, and synthesis. It does not perform inbox triage, tagging or metadata edits, exports, or archive maintenance unless the user explicitly overrides scope.
<quick_start>
command -v readwise
readwise --help
If authentication is missing, run readwise login or readwise login-with-token <token>.
readwise --json reader-search-documents --query "$TOPIC" --limit 20
readwise --json readwise-search-highlights --vector-search-term "$TOPIC" --limit 30
readwise --json reader-get-document-details --document-id "<document_id>"
readwise --json reader-get-document-highlights --document-id "<document_id>"
<process>.
</quick_start>reader-search-documentsreadwise-search-highlights
Use relevance-focused limits first (documents 20, highlights 30). Increase only when recall is weak.3: directly answers the research question with specific claim or detail2: supports a subtheme with useful context1: tangential but potentially useful0: irrelevant or noise
Keep score 2-3 items first. Use recency only as a tiebreaker.reader-get-document-details)reader-get-document-highlights)
Prefer exact highlighted text over paraphrase whenever available.score >= 2) or conflicting claims without sufficient support.
Perform up to 2 refinement rounds:Enforce section-level snippet caps. For each subtheme section, include at most 5 snippets. Select highest relevance first, then maximize source diversity.
Resolve conflicting evidence explicitly. When sources disagree:
---
title: "{{YYYY-MM-DD topic-slug}}"
created: "{{ISO-8601 datetime}}"
topic: "{{research topic}}"
source_system: "readwise-cli"
tags:
- research
- readwise
- obsidian
- pkm
---
# {{research topic}}
## Research Question
{{one-sentence question}}
## Key Findings
- {{finding 1}}
- {{finding 2}}
- {{finding 3}}
## Evidence by Theme
### {{theme 1}}
1. "{{verbatim snippet}}" — [{{source title}}]({{source url}}) _(author: {{author|unknown}}, category: {{category|unknown}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}})_
2. "{{verbatim snippet}}" — [{{source title}}]({{source url}}) _(author: {{author|unknown}}, category: {{category|unknown}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}})_
### {{theme 2}}
1. "{{verbatim snippet}}" — [{{source title}}]({{source url}}) _(author: {{author|unknown}}, category: {{category|unknown}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}})_
## Conflicting Evidence
- **View A:** {{claim summary}}
Evidence: [{{source title}}]({{source url}}) _(author: {{author|unknown}}, category: {{category|unknown}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}})_
- **View B:** {{claim summary}}
Evidence: [{{source title}}]({{source url}}) _(author: {{author|unknown}}, category: {{category|unknown}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}})_
- **Why they differ:** {{reason grounded in source context}}
## Confidence and Gaps
- **Confidence:** {{High|Medium|Low}} — {{justification}}
- **Known gaps:** {{missing or underrepresented evidence}}
- **Next queries:**
- {{query refinement 1}}
- {{query refinement 2}}
## Sources
- [{{source title}}]({{source url}}) — author: {{author|unknown}}, category: {{category|unknown}}, location: {{location|n/a}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}}
- [{{source title}}]({{source url}}) — author: {{author|unknown}}, category: {{category|unknown}}, location: {{location|n/a}}, document_id: {{document_id|n/a}}, highlight_id: {{highlight_id|n/a}}
url unavailable and keep metadataYYYY-MM-DD topic-slug.md<success_criteria> This skill is successful when all conditions are true:
Conflicting Evidence section with both views when disagreement exists.Confidence and Gaps section, even when confidence is high.