The interface is localized. Skill descriptions and documents are machine-translated by AI into your language; code, commands, paths and technical terms stay in the original, and the author's document governs.
Explore skills
Documents · SKILL.md

knowledge-synthesis

knowledge-synthesis

Synthesize information from multiple enterprise sources into a coherent, well-structured summary. Combines data from email, chat, docs, wikis, and tickets to produce unified knowledge artifacts that eliminate information silos and surface connections across tools. TRIGGER when: user asks to synthesize, combine, consolidate, or merge information from multipl…

Independent GitHub snapshot · Automated classification · Original documentation governs

Skill documentation

GitHub

name: knowledge-synthesis description: > Synthesize information from multiple enterprise sources into a coherent, well-structured summary. Combines data from email, chat, docs, wikis, and tickets to produce unified knowledge artifacts that eliminate information silos and surface connections across tools. TRIGGER when: user asks to synthesize, combine, consolidate, or merge information from multiple sources, create a knowledge summary, build a briefing document, or asks "what do we know about X". argument-hint: " [--sources SOURCE1,SOURCE2] [--depth shallow|deep]" user-invocable: true

Knowledge Synthesis

Combine and reconcile information scattered across enterprise tools into a single, authoritative summary with proper attribution and confidence scoring.

Synthesis Process

Step 1: Understand the Knowledge Request

Classify the request type:

Request TypeDescriptionDepth
Quick AnswerSimple factual questionShallow
Topic BriefingOverview of a subject or projectMedium
Decision SupportComprehensive analysis for a decisionDeep
Onboarding PackageEverything a new person needs to knowDeep
Competitive IntelAggregated info about a competitor/marketDeep
Incident PostmortemTimeline and facts about an eventDeep

Extract from the user's request:

  • Core topic or question being investigated
  • Scope boundaries — what to include and exclude
  • Target audience — who will consume this synthesis
  • Desired format — narrative, bullet points, Q&A, timeline

Step 2: Multi-Source Information Gathering

Query each relevant source with topic-appropriate search terms:

  1. Wiki / Knowledge Base — Canonical documentation, policies, architecture docs
  2. Documents — Specs, proposals, meeting notes, design docs
  3. Email — Decisions, approvals, context from stakeholder threads
  4. Chat / Messaging — Informal decisions, real-time discussions, shared links
  5. Tickets / Issues — Implementation details, bugs, feature requests
  6. Code Repositories — READMEs, comments, commit messages (if connected)

For each source, capture:

  • The specific information found
  • The date of the information
  • The author or contributor
  • The confidence level (verified, likely, uncertain)

Step 3: Reconcile and Cross-Reference

SituationResolution Strategy
Consistent across sourcesHigh confidence; cite multiple sources
Info only in one sourceMedium confidence; note single-source limitation
Conflicting informationFlag the conflict explicitly; show both versions + dates
Outdated in one, current in anotherPrefer the most recent; note the evolution
Informal vs. formal sourcesPrefer formal docs; note informal context if relevant

Step 4: Structure the Synthesis

Output Format

# Knowledge Synthesis: [Topic]
Generated: [date] | Sources: [count] sources across [count] tools
Confidence: [High/Medium/Low] — [reason]

## Executive Summary
[2-4 sentence overview of key findings]

## Key Facts
| Fact                     | Source          | Date       | Confidence |
|--------------------------|-----------------|------------|------------|
| [Fact 1]                 | [Source + link] | [date]     | High       |
| [Fact 2]                 | [Source + link] | [date]     | Medium     |

## Detailed Findings

### [Subtopic 1]
[Narrative synthesis of what is known]

**Sources:**
- [Source 1]: [what it contributed]
- [Source 2]: [what it contributed]

### [Subtopic 2]
[Narrative synthesis of what is known]

**Sources:**
- [Source 1]: [what it contributed]

## Conflicts and Ambiguities
- **[Topic]**: [Source A] says X (as of [date]), but [Source B] says Y (as of [date]).
  Recommended resolution: [suggestion]

## Knowledge Gaps
- [ ] [What is unknown or not found in any source]
- [ ] [Questions that remain unanswered]
- [ ] [Areas where information is stale and needs refresh]

## Timeline of Key Events
| Date       | Event                           | Source        |
|------------|---------------------------------|---------------|
| [date]     | [Event description]             | [Source]      |

## Source Index
| # | Source                | Type   | Last Updated | Items Used |
|---|----------------------|--------|--------------|------------|
| 1 | [Document name/link] | Doc    | [date]       | [count]    |
| 2 | [Channel name]       | Chat   | [date]       | [count]    |
| 3 | [Ticket ID]          | Ticket | [date]       | [count]    |

Synthesis Depth Levels

Shallow (Quick Answer)

  • Search top 5 results per source
  • Return the most authoritative single answer
  • Include 1-2 supporting citations
  • Target: under 200 words

Medium (Topic Briefing)

  • Search top 15 results per source
  • Cross-reference across 2-3 sources minimum
  • Identify key facts, people, and dates
  • Target: 300-800 words

Deep (Comprehensive Analysis)

  • Exhaustive search across all sources
  • Full cross-referencing and conflict resolution
  • Timeline construction and trend analysis
  • Knowledge gap identification
  • Target: 800-2000 words

Quality Standards

  • Attribution: Every claim must cite at least one source
  • Recency: Always note the date of information; flag anything older than 90 days
  • Neutrality: Present conflicting viewpoints without bias
  • Completeness: Explicitly state what was NOT found (knowledge gaps)
  • Traceability: Reader should be able to verify any claim by following the source link
  • No hallucination: Never infer or fabricate information not found in sources

Edge Cases

  • No results found: Report clearly that no information was found; suggest alternative search terms or sources to check manually
  • Single source only: Produce the synthesis but prominently note the single-source limitation
  • Highly conflicting sources: Escalate to the user before choosing a "winner"; present both sides
  • Sensitive or restricted content: If a source indicates restricted access, note the existence of the source without revealing restricted content
  • Stale information: If all sources are older than 6 months, add a staleness warning to the synthesis

Follow-Up Options

After delivering the synthesis, offer:

  1. "Want me to go deeper on any subtopic?"
  2. "Should I search for additional sources to fill the knowledge gaps?"
  3. "Want me to turn this into a shareable document or wiki page?"
  4. "Should I set up alerts for new information on this topic?"

Quality Checklist

  • Output is specific and actionable, not generic
  • All relevant inputs have been gathered before producing output
  • Recommendations are prioritized by impact
  • Stakeholders and audience are identified
  • Output format matches the audience's needs
  • Key assumptions are documented
  • Follow-up actions have clear owners
Saved skills