throughline — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited throughline (Plugin) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Small, sharp, agent-agnostic skills that keep an AI agent's work honest.
A whetstone doesn't make a tool — it keeps its edge true. Whetstone is a growing collection of tiny open-source skills for AI agents: not frameworks, but disciplines that keep an agent's work on-track, high-quality, and easy to hand off.
local model — and as a skill / slash command where those exist. Nothing assumes a particular platform; every capability degrades gracefully.
optional, zero-dependency, and never required.
light has to mean something.
that's a defect, not concision.
| Skill | What it does | Works with any AI / model |
|---|---|---|
| [Throughline](./handoff-skill/throughline) | Task-Track & Handoff for AI agents on long, multi-session work — within a session, one ⭐️ MAIN goal anchor in an emoji-tagged task list so reasoning doesn't drift; across sessions or agents, a plain-markdown handoff the next agent re-derives before trusting. | ✅ Yes — prompt-first, agent-agnostic |
| [Session Measurement](./session-measurement) | Score an agent session by session and read the trend across versions — so a model / frame / skill change shows up as a real regression or improvement, not a vibe. Six honest counts per session, plain-markdown trend table (optional charts), any agent. | ✅ Yes — any agent |
| [notify-me](./notify-me) | Calls you back when an agent needs you — the moment it gets blocked waiting on your decision, it pushes to your phone (via ntfy) and/or chimes on your computer, so you can step away and still know when you're needed. Hook-based, no account, no background process. | ✅ Yes — any agent with hooks |
| _more coming_ | — | — |
Whetstone is a Claude plugin marketplace, so you can install these skills directly in Claude Code or Cowork:
/plugin marketplace add jovesun-lab/whetstone
/plugin install session-measurement@whetstone
/plugin install throughline@whetstone
/plugin install notify-me@whetstoneInstalled skills are enabled by default and trigger automatically when relevant. Pull future updates with /plugin marketplace update whetstone. Once Whetstone is accepted into Anthropic's plugin directory, you'll also be able to find and install these from the in-app Plugins directory (Customize ▸ Plugins) without adding the marketplace by hand.
Each skill is a self-contained folder. Drop it where your tool reads skills / commands (Claude Code, Cursor, …), or just paste the relevant file into any agent as a prompt. Start with each skill's own README.md.
Each Whetstone skill is built to be cheap to keep installed and modest to use (per skill, approximate, tokenizer-dependent):
Throughline session that writes one handoff runs ~3–5k).
On progressive-disclosure hosts (Claude Code, Cursor) you pay the ~250 baseline and spend the rest only when a skill actually fires — and triggers are scoped so short, one-pass tasks don't load them.
Every Whetstone skill is MIT unless noted otherwise. Free to use, modify, embed, redistribute.
Whetstone is by [guxcom](https://guxcom.io). Feedback, issues, and PRs welcome on any skill.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.