8000 GitHub - keithmackay/mackayi: Keith MacKay's Claude Code plugin marketplace · GitHub
[go: up one dir, main page]

Skip to content
keithmackayPublic

About

Keith MacKay's Claude Code plugin marketplace

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

35 Commits

Folders and files

Repository files navigation

mackayi Marketplace - Skills & Plugins

Keith MacKay's Claude Code plugin marketplace. This repo hosts only the marketplace catalog (.claude-plugin/marketplace.json) — each listed plugin lives in its own repository and keeps its own version history, README, and release cadence.

Installation

/plugin marketplace add and /plugin install are slash commands — run them inside an active Claude Code session, not in your shell.

/plugin marketplace add keithmackay/mackayi
/plugin install <plugin-name>@mackayi

If the install summary says Run /reload-plugins to activate., run that command too.

For non-interactive installs, the marketplace must already be known — either added once via /plugin marketplace add in a prior session, or declared in .claude/settings.json under extraKnownMarketplaces — then install with the shell CLI:

claude plugin install <plugin-name>@mackayi --scope user

Plugins and Skills

The mackayi marketplace includes a variety of plugins and skills, spanning the following categories:

  • Project Lifecycle (tools to help create, release, and document skills/plugins/apps)
  • Improve & Refine (tools to radically improve your projects and comms clarity)
  • Product & Planning (product design and estimation skills)
  • Knowledge & Docs (Karpathy wiki with research tool, graph, and learning plan; obsidian tagger tool; pseudocode creator)
  • Monitoring & Optimization (track your session details automatically, get recommendations on how to use fewer tokens, and have a look back at a project for recommendations)
  • Marketing (URL scoring tool)
  • Utilities (mistral voxtral TTS tool)
  • Security (offline safety scan of a skill, plugin or MCP server before you install it)

These are grouped by purpose below (a README-level convention only — marketplace.json itself has no category field, so this grouping isn't visible in the /plugin UI, just here). Version numbers reflect each repo's latest GitHub release and are kept in sync automatically by /git-release.

Project Lifecycle

Plugin Version Repo
bootstrap v1.2.0 keithmackay/bootstrap
port-skill v1.1.0 keithmackay/port-skill
make-readme v1.2.0 keithmackay/make-readme
git-release v1.2.0 keithmackay/git-release

Improve & Refine

Plugin Version Repo
improve-this v1.1.0 keithmackay/improve-this
plsfix v1.2.0 keithmackay/plsfix
humanify v1.1.0 keithmackay/humanify

Product & Planning

Plugin Version Repo
marty-caganize v1.1.0 keithmackay/marty-caganize
estimate-acem-cost v1.1.0 keithmackay/estimate-acem-cost

Knowledge & Docs

Plugin Version Repo
wikify v1.1.0 keithmackay/wikifyskill
pseudocodify v1.1.0 keithmackay/pseudocodify
obstagger v1.3.0 keithmackay/obstagger

Monitoring & Optimization

Plugin Version Repo
sessionstats v1.1.0 keithmackay/sessionstats
tokentamer v1.6.0 keithmackay/tokentamer
do-retro v1.2.0 keithmackay/do-retro

Marketing

Plugin Version Repo
url-eval v1.1.0 keithmackay/url-eval

Utilities

Plugin Version Repo
voxtral-tts v1.0.0 keithmackay/voxtral-tts

Security

Plugin Version Repo
scanskillsafety v1.3.0 keithmackay/scanskillsafety-skill

Plugin Details

sessionstats

Auto-tracks Claude Code sessions: hooks fire on session start/end to record project name, timestamp, session ID, duration, and cost/token metrics parsed from the session transcript against a built-in pricing table. Data is stored per-project in .sessionstats/session_stats.json with a regenerated markdown view. Provides /session_stats (per-project summary) and /sessionstats_report (cross-project totals, filterable by tag), plus per-model breakdown, orphan/crash detection, tagging, and multi-user support. Runs automatically once installed — no manual invocation needed.

Help · Changelog

port-skill

Ports a skill/plugin/slash command between four coding-agent platforms (Claude Code, Codex, Antigravity, Gemini CLI) via /port-skill <path-to-skill-directory> [--dry-run]. Detects the source platform, generates the required manifest/context files for each target platform, adapts the SKILL.md content, and documents platform-specific feature gaps in a ## Platform Limitations section and compatibility matrix. Also rewrites the skill's README with per-platform install instructions and validates every generated file before finishing.

Help · Changelog

improve-this

An evaluate-only review skill that inspects a project and reports potential improvements without ever modifying files. Infers the project type, proposes tailored evaluation categories (UI/UX, code efficiency, test coverage, security for code; clarity, completeness, redundancy, token efficiency for docs/skills), then produces a priority-ranked, categorized list of findings with Impact/Confidence ratings. Can optionally save the report and generate a phased implementation plan.

Help · Changelog

url-eval

Scores candidate domain names/URLs across 8 weighted, research-backed dimensions (memorability, spelling reliability, pronunciation, brand fit, relevance, competitor overlap, TLD appropriateness, etc.) via /url-eval. Checks live availability via web search, generates and scores alternative name suggestions, and produces a structured report with top-3 recommendations and a full score table. An --update flag refreshes the baked-in scoring research via live web search.

Help · Changelog

plsfix

Improves spec documents, prompts, requirements docs, or any instruction set by applying 12 writing principles (Structure, Content, Delivery) synthesized from major AI-lab prompt-engineering guidance. Reads the document, flags principle violations with location, rewrites it with minimal changes (marking assumed details with [CONFIRM] tags), and produces a change report mapping each edit to its rationale.

Help · Changelog

humanify

Works on Claude Code, Codex, Antigravity, and Gemini CLI. Detects and rewrites the linguistic patterns that make AI-drafted prose read as machine-generated: banned clichés, AI-favored vocabulary, copula substitution, participle-ending sentences, elegant variation, and list-length uniformity. Edits are targeted redlines, not a rewrite; quotes, citations, code, URLs, and proper nouns are flagged but never changed. --dry-run reports findings with a findings-per-1,000-words density figure without editing, and --voice <file> calibrates house-style checks against a sample of the author's own writing.

Help · Changelog

estimate-acem-cost

(experimental, ymmv) Estimates what it would cost (time/dollars) for an AI coding agent to build an existing codebase, or forecasts token + human-review cost for a planned agentic project, using the ACEM model (Total_Cost = C_LLM + C_HITL + C_Infra). Inventories the codebase into artifact types, assigns token estimates, applies corrective multipliers, estimates human-in-the-loop review cost, and runs a bundled calculator script (optionally Monte Carlo) to produce a p10/p50/p90 cost range rather than a single point estimate.

Help · Changelog

wikify

Builds and maintains an LLM-compiled knowledge wiki following the "Karpathy pattern," where the LLM handles bookkeeping (page creation, cross-references, contradiction detection, confidence tracking) while the human curates source material in a raw/ directory. Auto-detects context on invocation, routing to Init, Ingest, Query, Lint, or Learning Plan sub-workflows based on directory state. A bundled script can render the wiki into a website; also builds a knowledge graph.

Help · Changelog

make-readme

Generates or improves a project's README.md by analyzing the codebase (package manifests, directory structure, CI config, entry points) to auto-detect project type, language, and framework. In Create mode, builds a full README from a section template with conditional sections; in Improve mode, maps and scores existing sections (Strong/Adequate/Weak/Missing), presents a gap report, and enhances only what's needed. Also offers to generate companion files (CONTRIBUTING.md, LICENSE, CHANGELOG.md, SECURITY.md, issue/PR templates) — including, for a project that's itself a skill or plugin, a --help/:help mechanism backed by a help.md file.

Help · Changelog

marty-caganize

Applies Marty Cagan/SVPG product-management methodology (INSPIRED, EMPOWERED, TRANSFORMED) to product strategy, team structure, discovery, and roadmap work. Provides frameworks including empowered vs. feature teams, the PM/design/tech-lead product trio, vision/strategy/roadmap distinctions, the four discovery risks with opportunity solution trees, and outcome-based OKR writing. Used both diagnostically (classifying a described team/artifact and flagging anti-patterns) and generatively (drafting vision docs, OKRs, discovery plans), including staged/partial adoption guidance for constrained orgs.

Help · Changelog

bootstrap

Starts a new software project from scratch: copies template files (including a seeded CHANGELOG.md), sets up README.md, initializes git, optionally creates a GitHub repo (public/private), then guides an idea-refinement design session and produces a full phased implementation plan written to docs/plans/. Safe to re-run on an already-bootstrapped project — detects existing setup and never overwrites or duplicates completed steps.

Help · Changelog

git-release

Prepares a project for public release on GitHub: adds an MIT license, verifies README.md, CHANGELOG.md, and .gitignore, checks for a --help/:help mechanism if the project is itself a skill or plugin (adds a --version/:version command directly if missing, which reports the installed version plus a best-effort check for a newer GitHub release), creates a GitHub remote if one doesn't exist, applies branch protection requiring PRs, bumps the version field in every plugin manifest to match the release tag, finalizes CHANGELOG.md by mechanically renaming its Unreleased section to the new version and date (refusing to proceed silently — and asking for confirmation instead — if Unreleased is empty; never generating changelog prose itself), offers to sync the version into any local marketplace listing, and cuts a tagged GitHub release.

Help · Changelog

pseudocodify

Wraps the pseudocodify CLI to convert a codebase into human-readable, language-agnostic pseudocode. Analyzes the codebase in two phases — building a structured map, then generating per-file pseudocode using that map for cross-file coherence — in CLRS/Cormen, Structured English, or Pascal-like style, with incremental re-runs and a Recursive Language Model fallback for codebases exceeding context limits. Useful as a first step before porting logic to a new language, or for documenting unfamiliar code without reading every file.

Help · Changelog

do-retro

Generates or updates a docs/PROJECT_HISTORY.md chronicling a project's development history: git commits, design decisions, and every user prompt pulled from Claude Code session transcripts, plus a retroactive-learning section of hindsight lessons. Supports flags to generate subsets into separate files (e.g. --prompts). Triggers on "build story," "document how this was built," "do a retro," or "do-retro."

Help · Changelog

tokentamer

Audits a project's Claude Code session transcripts to find concrete, evidence-backed opportunities to have used fewer tokens: repeated/duplicated work, context pollution, unused MCP servers, always-loaded skill listings, costly subagent delegation, poorly-disclosed skills, bloated prompts, verbose CLAUDE.md/memory files, wrong model choices, missed memory-save opportunities, and places a deterministic script would have beaten an LLM call. Every finding cites a session id, timestamp, and quote or tool-call sequence — a bundled script parses transcript JSONL directly (including subagent transcripts) rather than asking the model to eyeball raw logs. Analyzes transcript history, not the codebase itself.

Help · Changelog

voxtral-tts

Converts text to speech using Mistral's Voxtral model ($0.016/1,000 chars, ~70ms latency) and delivers it as a voice note over Discord or Telegram, or hands back a local file path. Triggers on an explicit --voice flag or phrases like "read that to me." Supports 8 voices, --strip-md to clean agent output before speaking it, and --truncate for overly long input. Requires a MISTRAL_API_KEY.

Help · Changelog

obstagger

Fills in YAML frontmatter (context, type, subtype, subsubtype, tags) on Obsidian notes — a single file or an entire folder tree recursively — guided by a living per-vault taxonomy schema file. Proposes and writes new taxonomy entries and tags to the schema automatically when content doesn't fit existing ones, and --updatetags can build a schema from scratch by analyzing a vault's existing notes. Never silently overwrites populated frontmatter fields unless --yolo is passed. Works with any vault, remembering the last one used as the default for next time.

Help · Changelog

scanskillsafety

Statically scans a skill, plugin, or MCP server for known prompt-injection, exfiltration, obfuscation, secret, and destructive-command patterns before it's installed, and reports a red/yellow/green rating with the file and line of every hit. Reads the target's own files (SKILL.md, plugin manifests, commands, agents, hooks, .mcp.json, and the scripts its skills and plugins bundle) from a GitHub/Gitea URL or a local path, fully offline apart from that host, and never runs the target's code. The same scanner produces the safety badge on findsafeskills listings. --deep checks the exact commit you would install: every file with no caps, what runs at install and start-up, its packages against OSV, how it compares with findsafeskills' scan, and, if you agree to the plan, Claude reading every file in a locked-down claude -p; it names every service it contacts before sending anything.

Help · Changelog

Adding a plugin

Each entry in .claude-plugin/marketplace.json's plugins array needs a name and a source. For a plugin in its own repo:

{
  "name": "plugin-name",
  "source": { "source": "github", "repo": "keithmackay/plugin-repo" },
  "description": "..."
}

The referenced repo needs its own .claude-plugin/plugin.json at its root (or wherever its source points), following the plugin manifest schema.

License

MIT © Keith MacKay

About

Keith MacKay's Claude Code plugin marketplace

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages

0