samgalanakis
- 16 skills
- 0 followers
- 6 hours ago last updated
- ▌ Pilcrow · samgalanakis bundleMake your clanker your editor. A prose linter, a set of editor commands anchored in classical style guides (Strunk, Williams, Zinsser, Pinker, Orwell, King), and project commands for voice capture and drafting. Use when reviewing, polishing, drafting, or auditing markdown, HTML, or plain-text prose. AI-tell detection is one feature among many.
- ▌ Cupel · samgalanakis bundleMake your clanker your investing analyst. A personal research companion that learns the user's edges (their job, their life, their circle of competence) and stress-tests ideas against the investing canon (Lynch, Mayer, Dorsey, Graham, Bogle, Housel, Bernstein, Marks). It reads and writes a single Obsidian-compatible "office" vault that accumulates the user's edges, trusted sources, watchlist, themes, positions, theses, and a decision journal. Use when the user wants to research a company, capture or develop an investment idea, scout for new opportunities beyond their current list, review a holding or their portfolio, decide how to allocate or rebalance, think through a buy/sell/hold decision, or get a status check on where they stand. It gives reasoned, mandate-grounded calls — likely scenarios, rough upside and timeframe, and the risks named — refusing only false precision dressed as fact and bare tip-bot signals, and never placing trades.
- ▌ Localref · samgalanakisClone a GitHub repo or download a reference source locally to /tmp/ for searching and reading. Use when you need to reference external code, docs, or examples to complete the current task.
- ▌ Moonshot · samgalanakisPush a feature past the conventional default: strip it to first principles, ask what world-class looks like for the underlying problem, hunt the frontier wherever it lives (research, other industries, legendary implementations, adjacent fields), and return ranked, implementable techniques the project has left untapped. Use when the user wants a feature to feel special instead of like every other app, asks 'what would the best version of this look like', or wants out-of-the-box approaches mined for a practical problem.
- ▌ Prospect · samgalanakisBenchmark your project against reputable peer projects and mine them for concrete improvements: survey your repo, clone references locally, extract enforceable mechanisms, audit your own code through each reference's written doctrine, verify every finding independently, and output a verified improvement batch. Use when the user wants to benchmark against comparable projects, mine external repos for practices, or run an outward-looking quality round.
- ▌ Wholehog · samgalanakisImplement the optimal end-state directly with a clean cutover: replace superseded functionality completely, remove dead code, and avoid shims, migrations, adapters, or interim steps.
- ▌ Ghmonitor · samgalanakisMonitor GitHub Actions workflow runs using gh CLI. Polls for progress, reports status at regular intervals until completion or failure. On failure, diagnoses the issue and fixes it.
- ▌ Groundhog · samgalanakisMine all local Claude/Codex session traces for recurring interaction patterns and friction, then present a concise numbered report of improvement opportunities (automation, AGENTS.md/CLAUDE.md rules, precommit hooks, skills, new tools, prompting patterns, or anything else that fits). Strictly read-only: it recommends and stops — no changes until the user picks recommendations by number. Use when the user wants to understand how they work with coding agents and what to improve.
- ▌ Schemasmash · samgalanakis bundleRun a read-only, agent-orchestrated audit of a codebase's data representations — in-memory types and data structures, database schema, and the serialized shapes between them — for states that are representable but invalid, multiple sources of truth for one fact, ordering or counts stored instead of derived, boolean-flag and nullable-field state machines, stringly-typed values, god objects and god tables, layer drift, and invariants the types or schema fail to enforce. The coordinator inventories every representation cluster as a coverage contract, fans out bounded read-only reviews, independently verifies every finding, and runs one coverage check before ranking. Use when the user wants a data-model or type-design review, says the domain model or schema has grown janky, is about to add a feature that touches core types or tables, or wants agent-written code checked before bad representations compound. Strictly read-only: no edits, migrations, tests, commits, or pushes.
- ▌ Clankermerge · samgalanakisControlled interactive rebase or merge. Auto-resolves obvious conflicts; groups the rest into feature-level decisions and asks the user one compact question at a time.
- ▌ Wholehog Loop · samgalanakisIteratively drive an implementation to a target quality bar. By default the loop uses wholehog-score and a numeric target, but the user may provide an alternate critique method or stopping criterion. The main agent orchestrates implementation and independent critique passes until the criterion is met or a real blocker stops progress.
- ▌ Wholehog Score · samgalanakisScore a specific implementation, design, plan, feature, document, or repo state against the optimal wholehog end-state. Use when the user wants a concise scored report of what is good, what needs improving, the trajectory since the last run, and the next steps in optimal order.
- ▌ Complexitysmash · samgalanakis bundleAudit a codebase for systemic complexity and simplification opportunities: duplicated business logic, unnecessary abstractions, mixed responsibilities, poor boundaries, pattern drift, indirection without payoff, and modules that should be merged, split, or deleted. Use when the user wants a full-project design pass or says the codebase feels over-engineered, inconsistent, hard to change, or harder than the problem requires. Prefer this over spring-cleaning when the issue is architectural or cross-cutting rather than a smaller cleanup pass.
- ▌ Spring Cleaning · samgalanakisFull codebase health pass in two phases — first delete dead things (unused code, deps, shims), then restructure what remains (dedup, abstractions, pattern normalization, efficiency).
- ▌ Performancesmash · samgalanakis bundleAudit a codebase or system for real performance bottlenecks and optimization opportunities across CPU, memory, allocation behavior, data access, caching, databases, network hops, payload size, batching, connection management, concurrency, queueing, and compression. Use when the user wants a serious performance pass, suspects the system is slow or wasteful, or wants guidance on where latency, throughput, or resource efficiency are being lost. Prefer this over ad-hoc micro-optimizing or style-level cleanup.
- ▌ Visual Explainer · samgalanakis bundleTurn a topic, plan, codebase, architecture, or feature into a single self-contained HTML visual explainer — wireframes, diagrams, annotated code, file maps, and before/after panels in one offline file. Use when prose alone would be a wall of text and the reader would understand it faster as a scannable visual document.