Claude Skills by tstapler

Expand Missing Topics Command

Capture instincts from this session into MEMORY.md and optionally into Logseq wiki

Fetch the Wikipedia thumbnail image for any topic and save it as a local asset in the Logseq wiki. Returns the markdown image reference to embed in a page.

Scan text to find technical terms, concepts, and topics mentioned in plain text that should be wiki-linked or have their own zettels

Knowledge Maintenance Orchestrator

Finds journal entries marked with [[Book Recommendation]], researches books, integrates with book-sync system, creates wiki pages, and removes labels after success

Analyzes a journal entry (explicit links and implicit topics), generates comprehensive zettels for missing/incomplete pages using research-backed synthesis

Finds journal entries marked with [[Needs Handy Plan]], generates comprehensive construction/house project plans with tools, parts, safety, and instructions, creates Logseq pages, and removes labels after success

Finds journal entries marked with [[Needs Research]], discovers and incorporates child topic pages, conducts research for projects or products, creates comprehensive zettels with hierarchical awareness, removes labels after success

Finds journal entries marked with [[Needs Synthesis]], delegates to knowledge-synthesis agent for comprehensive processing, discovers and integrates child topic pages, removes labels after success

Takes a journal entry, analyzes linked pages, and generates any missing or empty zettels using synthesize_knowledge process

Creates a comprehensive Zettelkasten note optimized for Logseq with proper interconnections and semantic structure

Synthesize Slack threads I participated in into the Logseq daily journal. Default mode walks backwards until finding a day already synthesized.

Analyze external content and create comprehensive daily synthesis notes that consolidate all learning into a single reviewable daily Zettel

Validates all [[wiki links]] and

Validates all [[wiki links]] and

Run any iterative analysis-and-fix workflow using minimal-context sub-agents that write full results to temp files and return only short structured summaries. Prevents context poisoning while looping until a condition is met. Use when a skill needs "run until clean", "repeat until all findings fixed", or "check → fix → verify" cycles at scale.

Optimize LinkedIn profile for searchability, recruiter visibility, and engagement

Design and document a port assignment strategy for local development environments. Use when setting up new projects, resolving port conflicts, or establishing team conventions for port usage across services.

Use this agent when you need to parse, filter, and analyze log files

Reference for writing and formatting Logseq wiki pages in Tyler's personal wiki. Covers bullet syntax, indentation, page properties, linking strategy, image embedding, filename rules, and the two main page styles. Apply whenever creating or editing pages in logseq/pages/ or logseq/journals/.

Legacy Confluence-only sync via the markdown-confluence CLI — publish

Generate beautiful, well-styled Mermaid diagrams. Use when asked to create architecture diagrams, flowcharts, sequence diagrams, state machines, ER schemas, or any visual diagram using Mermaid syntax. Produces professional output with correct syntax, appropriate diagram type selection, and polished styling.

Evaluate new Claude and Claude Code techniques, tools, features, prompting patterns, or workflow changes for adoption value. Use when encountering blog posts, release notes, tutorials, community tips, or configuration changes related to Claude and want to assess whether they fit the user's existing workflow (Logseq wiki, Python tools monorepo, skills library), align with Anthropic best practices, and are worth adopting. Produces structured evaluations with go/no-go recommendations and integra...

Master context engineering for AI agent systems. Use when designing agent architectures, debugging context failures, optimizing token usage, implementing memory systems, building multi-agent coordination, evaluating agent performance, or developing LLM-powered pipelines. Covers context fundamentals, degradation patterns, optimization techniques (compaction, masking, caching), compression strategies, memory architectures, multi-agent patterns, LLM-as-Judge evaluation, tool design, and project ...

Audit a SKILL.md file for missing cross-references to related skills, or map a task description to the skills that should be combined to handle it. Writes inline section callouts and "Related Skills" tables using the canonical pattern. Use when improving an existing skill, building a new one, or determining which skills to activate for a complex multi-domain task.

Select appropriate Claude model (Opus 4.5, Sonnet, Haiku) for agents, commands, or Task tool invocations based on task complexity, reasoning depth, and cost/speed requirements.

Harden an agent's persona/system-prompt file (SOUL.md-style: a git-tracked identity + process-gate document) against a concrete recurring failure, instead of doing a vague "improve the persona" pass. Mines a real incident, converts judgment-based process gates into low-judgment triggers, verifies every skill/resource name the persona references against what actually exists on disk, adds a cheap inline check reachable mid-edit, and adds a reuse check targeted at the mined failure. Language- an...

Apply prompting techniques when creating prompts, agents, commands, system instructions, or SKILL.md files. Use for XML tags, multishot examples, chain-of-thought, response prefilling, and Claude 4-specific patterns.

Apply systematic research methodology for multi-step research, fact-finding, web search, or verification tasks. Use when performing Brave Search, Puppeteer navigation, or synthesizing information from multiple sources.

Audit existing Claude Code skills/agents/commands for authoring-quality problems — oversized SKILL.md files that should use progressive disclosure, weak/generic descriptions, dangling symlinks, and plugin-vs-namespace-vs-flat-skill structure. Use when reviewing the skills library for quality, deciding whether a skill should be split or restructured, or asked "which skills need work" or "is this skill too big." Complements `skill:create` (new-skill guidance) and `meta-skill-effectiveness-audit...

Measure real skill/subagent/command usage and token cost from Claude Code session transcripts, score the highest-impact ones across evaluation dimensions (task success, token efficiency, instruction adherence, latency, drift), and hand the worst offenders to `meta-persona-hardening` as concrete, data-backed incidents instead of vague "improve this skill" guesses. Use when auditing a skills library for cost/quality, deciding which skill to harden next, or asked "which skills are we actually us...

Analyze Claude Settings and Suggest Global Allows

Report what's consuming your Claude Code context window — token estimates, loaded files, MCP servers

When stuck in a loop, step back and systematically research the root cause

Guide through creating a new purpose-built agent for specialized tasks with proper configuration

Guide through creating a new Claude Code slash command with proper hierarchy and location

Refactor CLAUDE.md to be concise with references to external documentation, minimizing token usage while maintaining prescriptive standards

Improve and enhance existing agents, commands, or prompts using expert prompt engineering principles

Select appropriate Claude model (Opus 4.5, Sonnet, Haiku) for agents,

I'm building a multi-agent system where agents keep duplicating work and their contexts get polluted with irrelevant information from other agents. How should I design the coordination to maintain context isolation? - [ ] Emphasizes context isolation principle ("Isolation prevents degradation") - [ ] Recommends sub-agents for context isolation, not role-play - [ ] Addresses cost implications (~15x single agent baseline) - [ ] Suggests partitioning work across sub-agents - [ ] May mention Writ...

Use this skill to query Google NotebookLM notebooks and auto-synthesize answers into Logseq. Trigger when: user says 'notebooklm', 'ask my notebook', 'ask my research', 'query my notebook', 'load YouTube channel into NotebookLM', or wants source-grounded answers from a specific knowledge base. Do NOT trigger for general web research — this is only for querying content the user has already uploaded to NotebookLM.

Compare multiple job offers side-by-side with total compensation analysis

Audit a public GitHub repository for tip-top shape — repo metadata (description, topics, social preview, team/CODEOWNERS access), comprehensive README (badges, quickstart, API docs, examples), developer UX as a library consumer or Claude Code plugin user, and CI/CD workflows that validate real usage. Use when you want to make an open-source project or plugin shine for contributors and consumers, or to check that public repos have discoverable descriptions/topics set.

Plan a complete outfit (or coordinated couple's outfits) for a specific event. Anchors on a known statement piece (e.g. a partner's dress), derives a complementary color palette, audits owned wardrobe, identifies gaps, and builds a Logseq wiki lookbook page with images, product links, and 'why it works' pairing rationale. TRIGGER when the user wants to plan what to wear to a wedding, party, formal event, or date — or wants to coordinate looks with a partner.

Apply terse output style for this session: - Answer directly — no preamble ("I'll", "Let me", "Sure,", "I'd be happy to") - No postamble ("Let me know if...", "Hope that helps") - Bullets > prose - Code > explanation — show it, don't narrate it - Skip rationale unless asked - Stop when the answer is complete

Apply detailed output style for this session: - Full explanation with rationale and context - Show alternatives considered and trade-offs - Examples where helpful - Explain the "why", not just the "what" - Prose where it aids understanding over bullets

Create a numbered ADR in docs/adr/ for a significant architectural choice

Comprehensive feature planning with software engineering principles and LLM-optimized task breakdown

Pick out the next highest priority bug and work on it