My custom claude code (hopefully agentsmd soon) plugins for when I've scoured the interwebs and all the existing CC marketplaces but couldn't find anything good enough.
Skill Claude CodeCodex
Emit one goal statement for the current work, hard-capped under 4000 characters and measured with wc -m, never estimated. Takes an optional focus hint like /compact ('focus on the auth path, ignore docs'); with no argument it derives the objective from the session's recent pivots plus any unfinished plan or task…
Skill Claude CodeCodex
Emit a paste-ready two-part handoff for a fresh session: a ## Goal statement hard-capped under 4000 characters (produced by the /goal skill), plus an unbounded ## Full prompt carrying cwd, ordered reading list, hard rules, pitfalls, and deliverables. Runs with or without a git repository — branch and PR facts are…
Skill Claude CodeCodex
Rebuild a plan from live ground truth — the artifacts themselves, tests, and (in a repository) git and gh — when the existing plan file no longer matches reality — items silently done, approach superseded, or facts drifted. Use when a plan has gone stale mid-effort or when resuming reveals contradictions. Re-derives…
Skill Claude CodeCodex
Pick up unfinished work cold in a fresh session by re-deriving state from live git, gh, the plan file, and TaskList — never from remembered or pasted claims. Use at the start of a new session that continues prior work, or when handed a resume prompt. Verifies what is actually merged, open, committed, and checked-off…
Skill Claude CodeCodex
Analyzes current session state without any cleanup. Full mode (default): resolves the active plan file, reads plan checklist + TaskList, gathers unfinished work/issues from conversation history, and emits a /handoff-built next-session prompt. Mid-session mode (/session-status mid): a fast plain-language 'done vs…
Skill Claude CodeCodex
Record triaged follow-up work where it will actually be seen — create the item in the issue tracker (Vikunja) or the incident system of record (Zammad), deduplicating first and reporting the created identifier. Use when a session produced work that outlives it, or when another skill has triaged items and needs them…
Skill Claude CodeCodex
Emit one paste-ready prompt for a local (weaker) LLM whose sole job is updating public and private documentation to match every technical change this session made — an exhaustive per-item changelog (file/component path, what changed technically, commit SHAs, PR URLs, why), routed between public repo docs and the…
Skill Claude CodeCodex
End-of-session handler that first checks whether the current session's plan is actually complete. If complete: run a quick retrospective, emit a forward-looking follow-up prompt, and — in a git repository — refresh the repo and clean gone branches. If incomplete: skip cleanup and emit ready-to-paste resume prompts so…
Skill Claude CodeCodex
Autonomous maintenance orchestrator that continuously finds and dispatches work.
Skill Claude CodeCodex
Route a task to the right model — a native Claude subagent, Codex, or local MLX — based on task type.
Skill Claude CodeCodex
Offload a bounded subtask to the shared model router — discover the live model menu from the router's published contract, pick the cheapest capable tier, place a bounded call, and report honestly when the router is unreachable.
Skill Claude CodeCodex
Fan a plan or diff out to independent model families — a cloud coding-agent CLI, a cloud reasoning-agent CLI, and a local model server — for adversarial review, plus the gotcha each type of invocation hits. Use when a second (or third) opinion on a plan/diff is wanted from genuinely independent models, when one…
Skill Claude CodeCodex
Choose among the hosted models the shared router serves — discover current ids and prices from the public catalog, self-enforce a spend budget the router does not meter for you, respect the free-tier prompt-logging caveat, and request onboarding of a model the router does not serve yet.
Skill Claude CodeCodex
Use when the session runs a premium top-tier/SOTA model (any vendor, current or future — the session's model IS the premium lead). Keeps premium reasoning on judgment (intent, architecture, tradeoffs, final review) and delegates checkable labor to cheaper agents or local/free LLMs.
Skill Claude CodeCodex
Use when writing or reviewing code and documentation.
Skill Claude CodeCodex
Use when performing formal code review on PRs.
Skill Claude CodeCodex
Permission requirements for GitHub Actions.
Skill Claude CodeCodex
Canonical security patterns for GitHub Actions workflows.
Skill Claude CodeCodex
DEPRECATED — permissions moved to nix-claude-code; edit data/permissions/.nix there instead of this repo's JSON.
Skill Claude CodeCodex
DEPRECATED — permission sync is retired; permissions now live in nix-claude-code and are governed by the auto-mode classifier.
Skill Claude CodeCodex
Validate all README files in the repository for required sections and installation code blocks, and score each one 0-6 on a quality rubric (purpose paragraph, quickstart, usage examples, license, length). Checks section presence using config from .readme-validator.yaml or sensible defaults. Badge URL reachability is…
Skill Claude CodeCodex
Use when managing branches, resolving merge conflicts, syncing with the default branch, or working with worktrees. Covers both trunk repos (main) and git-flow repos (develop) — see gh-cli-patterns (github-workflows) for default-branch detection.
Skill Claude CodeCodex
Use when creating PRs, linking issues, managing PR comments, or creating GitHub issues.
Skill Claude CodeCodex
Work with repositories using the git-flow-next branching model (default branch is develop, production branch is main). Covers branch creation, worktree setup, atomic commits, PR targeting, validation on develop, and mandatory end-of-session promotion to main.