caveman
claudeTalk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
YANG's personal toolkit: daily-ops skills + 4-tier observability layer + nested CLAUDE.md gap detection. Bring your own agents.
Plugin yang-toolkit from marketplace harness in YangHungTW/harness. Commands read from the repo's manifest — 7 days ago.
This plugin runs code on your machine — its hooks fire automatically on events like tool use. The exact commands are listed below; read them before installing.
curate-claude-mdAudit and (re)generate nested CLAUDE.md files across the current repo. Upper-level CLAUDE.md holds technical rules; sub-directory CLAUDE.md files hold business logic and domain invariants. Also serves as the FALLBACK generator for /yang-toolkit:claude-md-gaps when the official claude-md-management p
dashboardRead the current repo's `.claude/ledger.jsonl`, join it with the live git branch diff (this branch vs its base, including uncommitted changes), and render a paired artifact -- an interactive HTML view (timeline, feature kanban, per-agent tokens, loop economics [accept rate + cost-per-accepted-change
goThe single door into yang-toolkit -- hand it a task, a backlog, or a question and it picks the owning command and the smallest sufficient loop, then runs it. Replaces having to remember the plan-feature -> execute-plan -> ledger sequence yourself. TRIGGER on /go, or when the user hands over a unit o
todayAggregate today's work surface across sources -- GitHub PRs/notifications, each tracked client repo's `.claude/ledger.jsonl`, and any optional config-gated issue/Slack sources. TRIGGER on /today, "today", "今日", "morning brief", "每日簡報", or a standalone greeting when no other task is present.
weekCross-repo weekly report. Scans every harness-tracked client repo's `.claude/ledger.jsonl`, produces a markdown weekly recap grouped by client AND by feature. TRIGGER on /week, "週報", "weekly report".
workspace-initRegister (or refresh) the projects in a workspace directory (e.g. ~/Projects -- not a git repo, holds many projects) into .claude/workspace.json, so /yang-toolkit:go can route a plain-language request to the right project. Fully automatic -- scripts derive the structure, the agent names aliases, not
/yang-toolkit:claude-md-gapsReview pending nested-folder CLAUDE.md candidates, delegate generation to the official claude-md-management plugin, gate on user confirmation, then write the file and ledger the result.
/yang-toolkit:execute-planExecute an accepted plan artifact. Reads .claude/plans/<slug>.md, validates criteria, resolves depends_on, builds a /goal condition, picks orchestration (auto by default), delegates to the project's own workflow skills where they own a phase and otherwise to tdd-feature or feature-dev-tracked, and a
/yang-toolkit:feature-dev-trackedWrap /feature-dev with per-phase persistence. Each phase emits a decision doc; the final phase appends a ledger record.
/yang-toolkit:ledger-appendManually append one record to .claude/ledger.jsonl. For backfilling a missed session, correcting an automatic Stop-hook entry, or (--close) auto-closing a merged feature from gh PR state.
/yang-toolkit:loopPlan/ledger-aware in-session heartbeat. Each tick auto-discovers the next runnable plan, hands it to /yang-toolkit:execute-plan behind an objective acceptance-criteria gate, persists the outcome, and arms the next wake-up. Propose-only by default; --unattended opts into auto execution. Hard token-bu
/yang-toolkit:plan-featureDraft a reviewable plan artifact for a feature. Runs a parallel research fan-out (codebase patterns + project history/ledger + conditional recency-grounded external research via deep-research), then plan mode, emitting .claude/plans/<slug>.md with auto-generated, typed Memory References. Does NOT ex
/yang-toolkit:share-planRender a plan.md (or any plan-shaped markdown) into a clean, self-contained, shareable HTML document for a non-terminal colleague to read. Read-only snapshot; does NOT modify the plan. Pairs with /yang-toolkit:plan-feature.
/yang-toolkit:statusOne-screen overview of in-flight work -- current feature, plan statuses, recent ledger tail. --abandon closes out an in-flight feature in one step (state + ledger + plan).
/yang-toolkit:tdd-featureTDD-driven feature workflow. Either continues from an in-flight /yang-toolkit:feature-dev-tracked session, picks up an explicit slug, or starts fresh. Enforces red -> green -> refactor cycles, writes a tdd-cycles log, and finishes with the standard ledger summary.
${CLAUDE_PLUGIN_ROOT}/hooks/seed-git-exclude.sh
${CLAUDE_PLUGIN_ROOT}/hooks/workspace-check.sh
${CLAUDE_PLUGIN_ROOT}/hooks/log-tool-call.sh
${CLAUDE_PLUGIN_ROOT}/hooks/record-claude-md-candidate.sh
${CLAUDE_PLUGIN_ROOT}/hooks/test-parity-check.sh
${CLAUDE_PLUGIN_ROOT}/hooks/doc-parity-check.sh
${CLAUDE_PLUGIN_ROOT}/hooks/reset-current-agent.sh
${CLAUDE_PLUGIN_ROOT}/hooks/update-current-agent.sh
${CLAUDE_PLUGIN_ROOT}/hooks/execution-checkpoint-guard.sh
${CLAUDE_PLUGIN_ROOT}/hooks/append-ledger.sh
${CLAUDE_PLUGIN_ROOT}/hooks/notify-complete.sh
Talk like caveman. Cut filler. Keep technical details explicit.
22 skills · 1 command · 3 agents · hooks
Persistent memory for Claude Code: the next session starts where the last one left off.
29 skills · 1 MCP server · hooks
tasteskill 2.5.1: professional company websites at the figma.com and x.ai level, with one section per screen, real photographs, real product UI, effects kept for tech and gaming pages, and a real-browser checker and finish gate.
1 skill
Open-source AI job search that runs locally in your AI coding CLI: scan job portals, evaluate listings into a structured A-H report with a 1-5 score, tailor your CV, track applications
5 skills · hooks
Cross-session memory and token savings for coding agents.
6 skills · 1 agent · 1 MCP server · hooks
Package any website or local web build into a lightweight desktop app with Pake CLI.
1 skill
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.