From pai
Executes Fabric patterns like extract_wisdom, summarize, analyze_claims, and threat_modeling natively in Claude context without CLI spawning. For content processing, security analysis, and wisdom extraction.
npx claudepluginhub jeffh/claude-plugins --plugin paiThis skill uses the workspace's default tool permissions.
**Fabric patterns are just markdown prompts.** Instead of spawning `fabric -p pattern_name` for every task, Claude Code reads and applies patterns directly from `${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/`. This gives you:
tools/patterns/agility_story/system.mdtools/patterns/agility_story/user.mdtools/patterns/ai/system.mdtools/patterns/analyze_answers/README.mdtools/patterns/analyze_answers/system.mdtools/patterns/analyze_bill/system.mdtools/patterns/analyze_bill_short/system.mdtools/patterns/analyze_candidates/system.mdtools/patterns/analyze_candidates/user.mdtools/patterns/analyze_cfp_submission/system.mdtools/patterns/analyze_claims/system.mdtools/patterns/analyze_claims/user.mdtools/patterns/analyze_comments/system.mdtools/patterns/analyze_debate/system.mdtools/patterns/analyze_email_headers/system.mdtools/patterns/analyze_email_headers/user.mdtools/patterns/analyze_incident/system.mdtools/patterns/analyze_incident/user.mdtools/patterns/analyze_interviewer_techniques/system.mdtools/patterns/analyze_logs/system.mdGuides writing effective LLM system prompts using five layers: role, context, task, constraints, output. Covers role specificity, constraints, structured formats, and examples. Use for prompt engineering or AI configuration.
Documents reconstructed prompt architectures for agentic AI coding assistants like Claude Code, covering system prompts, multi-agent coordination, security classification, and memory hierarchies. Activates on agentic AI research queries.
Conducts multi-source parallel research using Claude WebSearch, Perplexity, Gemini; performs deep content analysis and extraction from difficult sites, YouTube via BrightData scraping and Fabric patterns. Activates on research or analysis requests.
Share bugs, ideas, or general feedback.
Fabric patterns are just markdown prompts. Instead of spawning fabric -p pattern_name for every task, Claude Code reads and applies patterns directly from ${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/. This gives you:
For any pattern-based processing:
${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/{pattern_name}/system.mdExamples:
User: "Extract wisdom from this transcript"
→ Read ${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/extract_wisdom/system.md
→ Apply pattern to content
→ Return structured output (IDEAS, INSIGHTS, QUOTES, etc.)
User: "Create a threat model for this API"
→ Read ${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/create_threat_model/system.md
→ Apply pattern to the API description
→ Return threat model
User: "Summarize this article"
→ Read ${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/summarize/system.md
→ Apply pattern to article
→ Return summary
Only use the fabric command for operations that require external services:
| Operation | Command | Why CLI Needed |
|---|---|---|
| YouTube transcripts | fabric -y "URL" | Downloads video, extracts transcript |
| Update patterns | fabric -U | Pulls from GitHub |
| List patterns | fabric -l | Quick reference |
For everything else, use native patterns.
create_threat_model - General threat modelingcreate_stride_threat_model - STRIDE methodologycreate_threat_scenarios - Threat scenario generationanalyze_threat_report - Threat report analysiscreate_sigma_rules - SIGMA detection ruleswrite_nuclei_template_rule - Nuclei scanner templateswrite_semgrep_rule - Semgrep static analysis rulessummarize - General summarizationcreate_5_sentence_summary - Ultra-concise summarysummarize_paper - Academic paper summarysummarize_meeting - Meeting notesyoutube_summary - Video summaryextract_wisdom - General wisdom extractionextract_insights - Key insightsextract_main_idea - Core messageextract_recommendations - Actionable recommendationsextract_alpha - High-value insightsanalyze_claims - Claim verificationanalyze_code - Code analysisanalyze_malware - Malware analysisanalyze_paper - Academic paper analysisanalyze_debate - Debate analysiscreate_prd - Product Requirements Documentcreate_design_document - Design documentationcreate_mermaid_visualization - Mermaid diagramswrite_essay - Essay writingcreate_report_finding - Security findingsimprove_writing - Writing enhancementimprove_prompt - Prompt engineeringreview_code - Code reviewhumanize - Humanize AI textRun the update script to sync latest patterns from upstream:
./tools/update-patterns.sh
This will:
fabric -U to fetch upstream updates${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/Requirements: fabric CLI must be installed (go install github.com/danielmiessler/fabric@latest)
Each pattern directory contains:
system.md - The main prompt/instructions (this is what gets applied)README.md - Documentation (optional)user.md - Example user input (optional)| Aspect | Native Patterns | fabric CLI |
|---|---|---|
| Model | Your subscription (Opus/Sonnet) | Fabric's configured model |
| Context | Full conversation history | Just the input |
| Speed | Instant (no process spawn) | ~1-2s CLI overhead |
| Integration | Seamless with Claude Code | External tool call |
The patterns are identical. The difference is execution context and model power.
See all available patterns:
ls ${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/
Or browse: ${CLAUDE_PLUGIN_ROOT}/skills/fabric/tools/patterns/{pattern_name}/system.md