Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/zigrivers/scaffold/work-beadsnpx skills add zigrivers/scaffold --skill work-beadsgit clone --depth 1 https://github.com/zigrivers/scaffoldWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00106 | $0.08399 |
| Opus 5 | $0.00053 | $0.04200 |
| Sonnet 5 | $0.00021 | $0.01680 |
| Haiku 4.5 | $0.00011 | $0.00840 |
Grade A, and why
work-beads scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 505 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Work Beads
Work the Beads queue with the ship loop. Multiple agents run this concurrently with no memory of each other — the loop is self-contained on purpose; do not skip steps.
The loop contract (memorize this):
set identity once, then repeat up to N times (one bead in flight per agent):
refresh view -> select ONE bead -> claim atomically, then validate (lost the
claim? next candidate; dup/conflict? cooldown-release + next) -> worktree
-> build (draft PR on first push; renew lease on each push)
-> verify (make check-affected) -> review (mmr, 3-round cap)
-> enqueue (make mq-enqueue) and move on -> daemon lands + closes the bead
batch end (budget spent, queue drained, or P0/blocker): report in the slots
The agent's finish line is a green gate + passing review + the PR ENQUEUED —
the merge-queue daemon batches, lands, and closes the bead; do not wait for the
merge or bd close an enqueued bead yourself. (No merge queue installed? Land
via the serialized merge slot instead — same finish line, you just do the
merge.) Standing authorization: run the whole loop without asking permission. Do
not end your turn after opening a draft PR with a list of "next steps" — that is
the #1 observed agent failure. The only mid-loop stops: a verified,
still-reproducing P0, or a blocker you can name.
Invocation: /work-beads (1 bead) · /work-beads N (up to N beads, selected
one at a time at claim time — N is a budget, not a reservation; never
pre-pick a list) · /work-beads N <label> (same, scoped to a label) ·
/work-beads <id> [<id>...] (explicit IDs, worked in dependency order).
Step 0 — Orient (once per batch, read-only, from the primary checkout)
The primary checkout is the first entry of git worktree list. Run:
bd ready && bd stats
gh pr list --state open # open + draft PRs = live registry of what others build
git worktree list
make doctor # wedged home base? make doctor-fix (unattended-safe)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 505 lines · 106 tokens per session scan A d0a81b05f94b
work-beads is a skill published in the GitHub repository zigrivers/scaffold (5 stars, last pushed 3d ago), licensed MIT. It adds 106 tokens to every session and 8,399 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
drogon-create-controller
生成 drogon 控制器代码（.h + .cc），支持 HttpSimpleController、HttpController、WebSocketController 三种类型。.
drogon-gen-advice
生成 drogon AOP Advice 代码（11 个内建切面之一），区分拦截型/观察型，含 SyncAdvice 短路。.
drogon-gen-cmake
生成 drogon 项目的 CMakeLists.txt，支持 ORM、Redis、WebSocket、C++20、协程等特性。.
drogon-gen-coroutine-handler
生成 drogon 协程 handler / 协程中间件 / 协程 ORM 调用，正确区分 Task/AsyncTask，强制裸 handler 参数按值传递。.
drogon-gen-lambda-handler
生成 drogon 现代 lambda 路由代码（app().registerHandler），含 {N} 路径参数绑定与 constraints 混传。.
drogon-gen-orm-crud
生成符合 drogon 约定的 ORM CRUD 代码，支持回调式和协程式，覆盖 PostgreSQL、MySQL、SQLite3。.