Plugin Claude Code
Fetch one or more recipes from URLs and save them as properly formatted Obsidian markdown notes in the user's vault. Use this skill whenever the user provides a recipe URL (or a list of URLs) and wants it saved to Obsidian — including requests like "save this recipe", "grab this recipe", "add to my cookbook", "fetch…
Skill Claude CodeCodex
Fetch one or more recipes from URLs and save them as properly formatted Obsidian markdown notes in the user's vault. Use this skill whenever the user provides a recipe URL (or a list of URLs) and wants it saved to Obsidian — including requests like "save this recipe", "grab this recipe", "add to my cookbook", "fetch…
Plugin Claude Code
The full self-improvement learning loop: autonomous background capture of corrections, errors, and gaps to /.learnings/ (the learnings skill), plus autonomous review that investigates each pending entry and auto-promotes the qualifiers into CLAUDE.md with a conservative bar and a revertible CHANGELOG trail. Review f.
Agent
Judges whether a single pending learning entry deserves to be promoted into CLAUDE.md. Reads one entry plus the target CLAUDE.md and returns a structured verdict (promote / confidence / scope / dedup / proposed text). Invoked by the self-improvement pipeline, one instance per candidate. The orchestrator launches these…
Agent
Proactive capture agent. Invoked in the background by main Claude whenever a trigger fires (user correction, command failure, feature gap, useful suggestion, knowledge gap). Writes a structured entry to /.learnings/ with a 6-char hex ID. Do NOT invoke this agent for normal coding tasks — only when something worth…
Agent
Runs the self-improvement review pipeline: sweep a session transcript for missed learnings, investigate each pending entry in /.learnings/ for 'needed-ness', and AUTO-PROMOTE the qualifiers into CLAUDE.md with a full revertible audit trail in CHANGELOG.md. Non-interactive — it never waits for approval (so it is safe…
Command
Undo a promotion that the self-improvement pipeline made. Usage: /self-improvement:revert (e.g. /self-improvement:revert PROMO-a3f7c1).
Command
Run the self-improvement review pipeline now, in the foreground, against this machine's pending learnings.
Command
Wire up the self-improvement plugin (capture + autonomous review) after installation or update.
Skill Claude CodeCodex
Autonomous capture of every correction, error, suggestion, and knowledge gap to /.learnings/ as it happens. Not user-invoked — this is the contract for passive logging. Main Claude delegates to the learnings subagent in the background the moment a trigger fires. Reviewing and promoting those entries into CLAUDE.md is…
Skill Claude CodeCodex
Review and AUTO-PROMOTE accumulated learnings into CLAUDE.md. Sweeps a session for missed learnings, investigates each pending entry in /.learnings/ for "needed-ness", and promotes the qualifiers into CLAUDE.md with a conservative bar and a full revertible audit trail in CHANGELOG.md. Runs autonomously when a session…
Plugin Claude Code
Launch background subagent(s) to handle a task as a side quest while the main conversation stays free. Use when the user types /side-quest or asks to run a task as a side quest. Rates the task's difficulty (Challenge Rating) to pick the model — haiku for trivial quests, sonnet for standard quests, opus for boss fights.
Agent
First-time setup agent for the side-quest plugin. Deploys xp.sh, xpcore.py, statusline-command.sh, and the xp-sync-daemon.sh MQTT listener to their stable paths; merges the statusLine block, Stop hook, and PostToolUse tick hook into /.claude/settings.json; installs the sync daemon as a launchd (macOS) or systemd…
Command
Reset the side-quest XP ledger to a specific total. Pool-wide and destructive — every machine on the shared MQTT sync adopts the new total.
Command
Take the task in $ARGUMENTS and run it as a side quest: hand it off to one or more background subagents so this conversation stays free, then report in Dungeons & Dragons style — always paired with a plain-English summary.
Skill Claude CodeCodex
Launch background subagent(s) to handle a task as a side quest while the main conversation stays free. Use when the user types /side-quest or asks to run a task as a side quest. Rates the task's difficulty (Challenge Rating) to pick the model — haiku for trivial quests, sonnet for standard quests, opus for boss fights…
Plugin Claude Code
Install, remove, list, find, and update Claude Code skills using the npx skills CLI. Use this skill whenever the user wants to manage their agent skills — install a new skill, search the skills registry, remove a skill, check for updates, or update all skills. Also trigger for requests like "find me a skill for X", ".
Skill Claude CodeCodex
Install, remove, list, find, and update Claude Code skills using the npx skills CLI. Use this skill whenever the user wants to manage their agent skills — install a new skill, search the skills registry, remove a skill, check for updates, or update all skills. Also trigger for requests like "find me a skill for X"…
Plugin Claude Code
Expert guide for managing Tailscale tailnet policy files (ACLs). Covers the HuJSON policy file format, all policy sections (acls, groups, tags, SSH rules, autoApprovers, tests), the Tailscale API for validating and applying policies, and GitOps automation via GitHub Actions so changes are automatically validated on…
Agent
Sets up GitOps automation for a Tailscale policy repository — creates the GitHub Actions workflow, configures the correct auth method, and optionally adds a pre-commit hook. Use this agent when the user wants to manage their Tailscale policy in git, set up CI/CD for policy validation and deployment, or migrate from…
Agent
Reviews a Tailscale HuJSON policy file for security issues, overly broad rules, missing tests, and correctness. Use this agent when the user wants a second opinion on a policy, wants to audit an existing policy, or asks "is this safe?". The agent produces a structured report with findings by severity and concrete…
Agent
Drafts, edits, and updates Tailscale HuJSON policy files. Use this agent when the user wants to write a new policy from scratch, add or modify ACL/grant rules, add groups or tags, update SSH rules, or convert existing acls to the grants syntax. The agent produces a complete, valid policy file with inline comments and…
Skill Claude CodeCodex
Expert guide for managing Tailscale tailnet policy files. Covers the HuJSON policy file format, all policy sections (grants, acls, groups, tags, postures, SSH rules, autoApprovers, nodeAttrs, tests), the Tailscale API for validating and applying policies, and GitOps automation via GitHub Actions. Grants are now GA and…
Plugin Claude Code
Recovery strategy when the Write tool fails to create a new file. Use this skill whenever a Write or Edit tool call fails with an error related to creating a new file — such as missing parent directories, permission issues, or "file not found" errors on files that don't exist yet. This skill does NOT apply to editing…