skills

Personal Agent Skills by @vdsmon, packaged as a plugin marketplace. Most run on any SKILL.md-native host, and a few are Claude-Code-specific.

Each skill ships as its own plugin so you can install only what you want.

Naming convention

Plugins prefixed with cc- are Claude-Code-specific: they use features (SessionStart hooks, !`cmd` dynamic context injection, ${CLAUDE_SKILL_DIR}, context: fork) that don't exist on other Agent Skills hosts.

Plugins without the cc- prefix are portable. They follow the open Agent Skills format and work on any SKILL.md-native host: Claude Code, OpenAI Codex CLI, Gemini CLI, Cursor, Goose, OpenCode, Copilot, Amp, Roo Code, and many more.

The Plugins table below lists every plugin and its host: Host: CC only is a cc- plugin, Host: any is portable.

Install on Claude Code

Register the marketplace once:

/plugin marketplace add vdsmon/skills

Install any plugin by name (see the Plugins table for the full list):

/plugin install <name>@vdsmon-skills

Or run /plugins and pick from the interactive browser.

Install on OpenAI Codex CLI

This repo ships a native Codex plugin marketplace (.agents/plugins/marketplace.json), so you install the same way Claude Code does: register once, then pick plugins. Only the portable (non-cc-) plugins are listed, since the cc- plugins rely on Claude-Code-only features and won't run on Codex.

codex plugin marketplace add vdsmon/skills

Then browse and install from the interactive picker:

/plugins

Select a plugin and choose Install plugin (Space toggles enabled state).

Install on other hosts (portable plugins only)

Each host discovers skills in its own directory. Clone this repo, then drop the skill folder into the target path.

git clone https://github.com/vdsmon/skills
cd skills

Gemini CLI

mkdir -p ~/.gemini/skills
cp -r plugins/humanize/skills/humanize        ~/.gemini/skills/
# ... etc

Cursor

# Cursor reads skills from its configured skills directory; see Cursor docs.
cp -r plugins/humanize/skills/humanize <cursor-skills-dir>/

Goose, OpenCode, Roo Code, Copilot, Amp, etc.

Same pattern: cp -r plugins/<name>/skills/<name> into the host's skills directory. Per-host paths: agentskills.io/clients.

Universal installers

For one-command multi-host install:

• skillport: pip install skillport, targets many hosts via CLI or MCP.
• openskills: npm i -g openskills.
• agent-skill-creator: auto-converts SKILL.md to host-specific formats (.mdc, .md rules, etc.) for Cursor/Windsurf/Cline.

Plugins

Generated from .claude-plugin/marketplace.json (the source of truth) by scripts/sync-codex.sh, so do not hand-edit between the markers. Host: any is portable, Host: CC only needs Claude Code. Run /plugins for the full descriptions and triggers.

Plugin	Host	What it does
investigate	any	Investigate a reported error or incident end to end across every reachable system, never inferring past a missing source and instead stopping to raise to the human for access or clarity.
slack-draft	any	Draft a Slack message for the user to send: Slack mrkdwn, lead-with-conclusion, backticked identifiers and domain values, ASCII punctuation.
skill-polish	any	Post-mortem for any skill: scans a session for friction and applies concrete edits to the responsible skill file. Portable across SKILL.md-native hosts.
cc-tokenomics	CC only	Token usage, cache hit rates, and Max plan consumption analyzer.
cc-cache-keepalive	CC only	Keeps Claude Code's prompt cache warm on Max plans via a silent cron scheduled at session startup (skips resume/compact, dedup-guarded via CronList).
cc-usage-guard	CC only	Pause-at-limit guard for Claude Code.
prep-compact	any	Audit in-flight state before any context-compacting step, and produce a copy-paste /compact message plus a queue-able follow-up that chains the next action when compact finishes.
prep-goal	any	Interrogate a rough objective into a tight, verifiable /goal completion condition before handing it to an autonomous goal loop.
humanize	any	Strip AI-writing tells from text. Detects em-dash overuse, AI vocabulary, inflated significance, rule-of-three, sycophancy, and 20+ more patterns.
loop-finder	any	Loop discovery + race + feature-driven iteration.
brainstorming	any	Design-before-code gate. Explores intent, proposes 2-3 approaches, presents a design, and gets approval before any implementation.
systematic-debugging	any	Four-phase debugging discipline (root-cause investigation, pattern analysis, hypothesis, single-fix implementation).
skill-smith	any	Forge for Agent Skills: create, test, evaluate, optimize triggering, and package skills.
git-cleanup	any	Clean up stale git branches and worktrees.
strip-migration-cruft	any	Scan a repo for transitional / migration / phase / wave / story / legacy-alias cruft comments, bucket into safe-to-strip vs keep-semantic, propose surgical edits and execute after confirmation.
grilling	any	Relentless one-question-at-a-time interview that stress-tests a plan or design to convergence: facts get looked up in the codebase, decisions go to the human, and nothing is enacted until shared understanding is confirmed.
wayfinder	any	Chart work too big for one agent session as a shared map of investigation tickets on the repo's issue tracker: name the destination, sketch the fog of war, then resolve one ticket per session (grilling, research, prototype, task) until the way is clear.
teach	any	Stateful, multi-session teaching workspace: grounds every lesson in a MISSION.md, gathers trusted RESOURCES.md, produces short self-contained HTML lessons in the learner's zone of proximal development, tracks progress via learning-records, and builds storage strength through retrieval, spacing, and interleaving.
codebase-design	any	Shared vocabulary for designing deep modules: a lot of behaviour behind a small interface, placed at a clean seam, testable through that interface.

Layout

.claude-plugin/
  marketplace.json                       # Claude Code marketplace (source of truth)
.agents/plugins/
  marketplace.json                       # Codex CLI marketplace (generated, non-cc- only)
mise.toml                                 # Maintainer task runner: mise run sync | bump | verify
scripts/
  bump-plugin.sh                          # Version bump + marketplace sync
  sync-codex.sh                           # Rebuild Codex symlinks + marketplace + README table from the Claude side
plugins/
  skill-polish/                           # portable
    .claude-plugin/plugin.json
    .codex-plugin -> .claude-plugin       # symlink; Codex reads .codex-plugin/plugin.json
    skills/skill-polish/SKILL.md
  cc-tokenomics/                          # cc- = Claude Code only, NO .codex-plugin symlink
    .claude-plugin/plugin.json
    skills/cc-tokenomics/
      SKILL.md
      scripts/token-report.py
      reference/{economics,experiments,keepalive}.md
  cc-cache-keepalive/                     # cc- = Claude Code only, NO .codex-plugin symlink
    .claude-plugin/plugin.json           # Declares SessionStart hook
    hooks/keepalive.sh                   # Opt-in, flag-gated
    scripts/keepalive-noop.sh
  humanize/                               # portable
    .claude-plugin/plugin.json
    .codex-plugin -> .claude-plugin
    skills/humanize/SKILL.md

Both marketplaces share one source of truth: you author .claude-plugin/*, then run scripts/sync-codex.sh to regenerate the .codex-plugin symlinks, the Codex marketplace, and the Plugins table above. The symlink means each plugin has exactly one plugin.json, so versions never drift between hosts.

License

MIT. See LICENSE.