Claude Skills by tstapler

Format CVs for academic positions with publications, grants, and teaching

Systematically address all open GitHub PR review comments — fix code

I need to build an AI assistant that remembers user preferences, past conversations, and learned patterns across sessions. Design a memory architecture that balances retrieval accuracy with token efficiency. - [ ] Distinguishes between in-context and external memory - [ ] Recommends cross-session persistence strategies - [ ] Addresses retrieval/selection mechanisms (Select bucket) - [ ] Mentions knowledge graphs or structured memory - [ ] Balances memory retrieval with token budget - [ ] Sugg...

Use this skill when setting up or explaining Android versionCode/versionName for KMP or standard Android projects. Implements packed semver: major*1_000_000 + minor*1_000 + patch as a single integer that preserves semver ordering.

Apply software architecture best practices when designing or reviewing

Use ast-grep (sg) for semantic, syntax-aware code searching. Prefer over

Reference for which construct to build when the user asks to automate, remember, or make reusable something — a skill, subagent, hook, rule, CLAUDE.md entry, shell alias, stapler-squad workflow, backlog item, approval rule, or scheduled job. Use whenever a request has the shape "set this up so I/you can do X easily/automatically/every time/from now on" and it's not obvious which mechanism fits.

My Claude agent is hitting context limits after 10-15 exchanges. What strategies should I implement to extend conversation length while maintaining response quality? - [ ] Explains token utilization thresholds (70% warning, 80% trigger) - [ ] Recommends compaction/summarization strategy - [ ] Mentions the four-bucket strategy (Write, Select, Compress, Isolate) - [ ] Suggests progressive disclosure or just-in-time loading - [ ] Warns against placing critical info in middle positions

Use when setting up or reviewing Bazel CI on GitHub Actions. Covers

Use this when configuring, debugging, or extending a Bazel build for a Kotlin JVM project: - Setting up `MODULE.bazel` from scratch - Adding or updating `rules_kotlin`, `rules_jvm_external`, or Gazelle - Registering Kotlin toolchains - Managing the Maven lockfile (`maven_install.json`) - Generating BUILD files with Gazelle

Find the slowest parts of a SteleKit Bazel build/test run and CI job, and know which fixes actually move the needle. Use when asked to speed up Bazel builds/tests, investigate why CI is slow, or find build hotspots — locally or from a GitHub Actions run.

Guidance for migrating Bazel projects to newer versions, specifically

Discover new Claude models available in AWS Bedrock and add them to the claude-proxy configuration.

Scan ~/Documents/personal-wiki (Logseq pages and journals) for blog post candidates, manage a content pipeline backlog, and move ideas through stages from candidate to published post on the personal-website Hugo blog.

Profile Chrome/JavaScript/React apps to diagnose slowness. Covers triage (load vs interaction vs scroll), Playwright-based baseline capture, React Profiler component, Chrome Performance panel interpretation, bundle analysis, memory leak detection, fix verification, and programmatic large-trace analysis via Perfetto's trace_processor (the pprof-equivalent for Chrome traces). Invoke when a browser or React app is noticeably slow and you need to find and fix the bottleneck, or when you have a do...

Translate skills from one industry to another, identify transferable skills

Evaluate new Claude prompting patterns, tools, or workflow changes. Produces go/no-go recommendations and integration plans.

Research, find, and compare specific clothing items (blazers, trousers, shirts, shoes, ties, pocket squares) across multiple retailers at a target price point, style, and size. Validates product availability and sources working images. Outputs a comparison table ready for a wiki page. TRIGGER when the user needs to find a specific type of clothing item, compare options across brands, or fill a gap identified during outfit planning.

Compose Multiplatform UI testing guide for SteleKit. Covers the two-layer testing strategy: createComposeRule-based interaction tests (click, assert, type) using ComposeUITestBase, and Roborazzi screenshot regression tests. Use when writing new Compose UI tests, adding screenshot tests, or understanding how the existing jvmTest UI test infrastructure works.

Systematically analyze external codebases by cloning git repos or extracting archives to temp dirs, then reverse-engineering the code to extract functional/non-functional requirements, identify architectural patterns, document design choices, and catalog problems worth solving. Use when asked to analyze, understand, reverse-engineer, or evaluate an unfamiliar codebase.

Apply software architecture best practices when designing or reviewing systems, classes, modules, or services. Use when structuring new code, evaluating design decisions, applying SOLID principles, Clean Architecture, Hexagonal Architecture, or Domain-Driven Design patterns. Works across languages — includes specific guidance for Python and Java/Spring Boot.

Use ast-grep (sg) for semantic, syntax-aware code searching. Prefer over Grep for code pattern searches that depend on structure rather than text matching. Use when searching for function calls, class definitions, import statements, or any pattern where syntax context matters.

Systematic debugging frameworks for finding and fixing bugs - includes root cause analysis, defense-in-depth validation, and verification protocols

Idiomatic review for go-git/go-git v5 concurrent access. Use when reviewing Go code that imports go-git, uses git.Repository, Worktree, CommitIter, or ObjectStorage. Covers the library's documented non-thread-safety, per-repo mutex requirements, and iterator lifetime rules.

Apply Hexagonal Architecture, Clean Architecture, and Domain-Driven Design to Go package structure. Use when designing or reviewing a Go service's package layout, deciding how to split domain/application/adapter code, choosing between package-by-layer and package-by-feature, or right-sizing architecture for a Go project's actual complexity. Complements `code-architecture-best-practices` (language-agnostic principles) with Go-specific structural decisions.

Use gritql (grit) for AST-based multi-file code transformations. Use when renaming methods/classes, migrating APIs, or modernizing patterns across a codebase. Always preview before applying. Pairs with ast-grep for search-then-transform workflows.

Ground architecture/refactoring reviews in tool-generated evidence instead of code-reading alone — static coupling graphs (package dependency, call graph, struct/interface size) plus temporal coupling (files that change together in git history, independent of imports) combined into a complexity × churn hotspot score. The open-source technique behind CodeScene. Use before `architecture-review`/`find-refactor-candidates` to find WHERE to look; use those commands to analyze WHY once you're there.

Discover Java API signatures from compiled JARs using javap instead of guessing or relying on incomplete documentation. Use when encountering unknown methods, pagination patterns, union types, or compilation errors from incorrect API assumptions.

Idiomatic review for Kotlin Multiplatform (KMP) with Kotlin 2.x — expect/actual discipline, Swift boundary safety, ViewModel lifecycle, SQLDelight conventions

Bootstrap a new personal/side project using Tyler's researched default stacks — a full hosted web app (Angular, Rust+Axum, Connect-RPC/WebSocket, GCP Cloud Run, Neon+R2, OpenTofu), a library/CLI/MCP tool (Rust, clap, rmcp, cargo-dist, git-cliff, Homebrew tap — modeled on tstapler/kibitzer), or a local single-user app/dashboard (roll-your-own Axum+SQLite+static-SPA bolted onto an existing binary by default; PocketBase/TrailBase/Tauri only for specific standalone-app or native-desktop cases). S...

Trace static call paths through a codebase — forward ("what does this entry point reach") and backward ("what reaches this function") — for impact analysis before a change, blast-radius estimation for an incident, dead-code detection, or verifying a security-relevant path (auth check, sanitizer) actually sits on every route to a sink. Language-agnostic: prefers LSP-based call hierarchy (serena MCP) when available, falls back to per-language call-graph tools, falls back further to an ast-grep/...

Trigger when the user asks to reverse engineer a closed-source Qt5/C++ Windows PE binary on Linux, recover Qt meta-objects (classes, signals, slots), trace Win32 API calls via Wine relay or WineDbg, hook network send/recv with Frida, capture and parse proprietary binary protocols (Wireshark → ImHex → Kaitai Struct), or analyze unknown binary file formats from embedded hardware. Also triggers for: "what does RayStudio send over the network", "recover class names from this DLL", "decode this sc...

Execute Arlo Belshee's Core 6 Refactorings mechanically using available CLI tools across TypeScript, Go, Rust, and Java. Evaluates tool availability, routes to the most precise tool, and produces a reviewable git diff. Use for: rename symbol, extract function, inline variable, introduce variable, change signature, move symbol. Never hand-edits at scale. Covers: rename, extract, inline, introduce-variable, change-signature, move. Triggers on: Core 6, Belshee, mechanical refactor, rename method...

Orchestrate large structural code refactors combining semantic search (ast-grep) with AST-based transformation (gritql). Use for multi-file renames, API migrations, and pattern modernization with mandatory quality gates.

Trigger when the user asks to reverse engineer a closed-source Windows PE binary, DLL, COM server, or unknown binary format on Linux — where the target is NOT known to be Qt5. Use for: "analyze this DLL", "what does this EXE do", "decode this binary protocol", "trace Windows API calls", "find the network protocol", "reverse this file format", "hook send/recv on this process". For confirmed Qt5 targets, use code-re-qt5 instead.

Use when receiving code review feedback (especially if unclear or technically questionable), when completing tasks or major features requiring review before proceeding, or before making any completion/success claims. Covers four practices — receiving feedback with technical rigor, requesting reviews via code-reviewer subagent, verification gates requiring evidence before claims, and security/API-compat awareness. Essential for subagent-driven development, pull requests, and preventing false c...

Systematic investigation of errors, failures, and unexpected behaviors by searching personal-wiki history and external documentation. Use when debugging errors with stack traces, investigating incidents/outages, finding historical context for similar issues, analyzing recurring problems, or searching for past solutions. Combines Logseq knowledge base (journals, pages, incident notes) with web search for external docs. Follows structured methodology to extract error signatures, search historic...

Apply Spring Boot and Java coding standards when writing, reviewing, or designing Java code. Use for Spring Boot project structure, DDD patterns (Entities, Aggregates, Repositories, Domain Services), Clean Code principles in Java, PoEAA patterns (Repository, Data Mapper, Service Layer, Unit of Work), testing with JUnit 5 and Mockito, and Spring-specific best practices. Covers constructor injection, transaction boundaries, layer responsibilities, and common anti-patterns to avoid.

Best practices for the AWS Strands Agents SDK — structuring prompts, multi-agent patterns, structured I/O, and splitting monolithic agents into specialists. Use when designing or refactoring Strands-based agent systems.

Run parallel minimal-context checks across all failure dimensions, fix everything found, and repeat until all pass or max iterations reached.

Systematically implement a new feature following research-backed best practices from the Six-Phase Software Implementation Methodology, with intelligent parallelization and multi-agent coordination

Shipping readiness gate — parallel swarm across plan compliance, architecture, code quality, tests, security, product/UX, and operational readiness. Produces a GO / HOLD / FIX-THEN-SHIP verdict with a mandatory auto-fix loop.

Parallelization Examples and Patterns

Refactor code following principles, patterns, and best practices from respected literature

Comprehensive code review — 5 parallel specialized agents (testing, code quality, architecture, database, security) → adversarial skeptic pass to filter false positives → severity-labeled findings [BLOCKER]/[CRITICAL]/[MAJOR]/[NIT] capped for actionability.

Write Confluence-compatible markdown for publishing via markdown-confluence tool. Covers syntax, frontmatter, directory structure, and Confluence-specific features.

Analyzes a Claude Code session transcript (real on-disk JSONL, not `/context` output) to show where context tokens actually went, then gives concrete recommendations for cutting them.

Master context engineering for AI agent systems. Use when designing agent

Generates a self-contained HTML artifact showing Claude Code context window usage as a Sankey diagram flowing from total capacity → categories → top sub-items. ---

Create personalized, compelling cover letters from resume and job description