Waku Agent is a local-first personal AI assistant whose readable code implements the agent loop, memory, and evaluation system. It is for people who want an assistant they can run and understand on their own laptop, with memory stored in SQLite and built-in testing. The catalogue add-ons support its agent workflow.
Borrowing it
Nothing to install: this file belongs to ShenSeanChen/waku-agent. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ShenSeanChen/waku-agent/main/.claude/skills/ship/SKILL.mdgit clone --depth 1 https://github.com/ShenSeanChen/waku-agentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/shenseanchen/waku-agent/ship)<a href="https://agentmods.dev/skills/shenseanchen/waku-agent/ship"><img src="https://agentmods.dev/badge/skills/shenseanchen/waku-agent/ship/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/shenseanchen/waku-agent/ship"><img src="https://agentmods.dev/badge/skills/shenseanchen/waku-agent/ship.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Tool Misuse · line 21 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
What 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 5.1 | $0.00046 | $0.00668 |
| Opus 5 | $0.00023 | $0.00334 |
| Sonnet 5 | $0.00009 | $0.00134 |
| Haiku 4.5 | $0.00005 | $0.00067 |
Grade A, and why
ship 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 10d 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 — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Procedure
-
make lint— fix anything it flags before continuing. -
make gate— the release gate. Deterministic evals must pass 100%; judge evals must clear threshold when an API key is present. If the gate closes, fix the cause (or, if a finding is a genuine behavior change, update the eval WITH the user's agreement) — never push a red gate. -
Review
git statusandgit diff— confirm nothing unintended is staged (especially nothing from.waku/or.env). -
Commit with a detailed message:
- Subject: imperative, specific ("Fix triple-booking from first live test"), never generic ("update code", "fixes").
- Body: WHY the change exists, what evidence motivated it (live bug, trace, eval failure), and what verification it survived.
- End with:
Co-Authored-By: Claude <the current model's attribution line>
-
Push via a branch and a PR —
git push origin mainwill be REJECTED. Since 2026-07-26mainrequires theskills-and-evalscheck to be green BEFORE a commit can land, andenforce_adminsis on, so this applies to Sean and to Claude equally. There is no bypass and you should not look for one.git checkout -b <short-topic-branch> git push -u origin HEAD gh pr create --fill # body already written in step 4 gh pr checks --watch # ~30s gh pr merge --squash --delete-branch # ONLY Sean's own work, see below git checkout main && git pull -qThe rejection looks like
GH006: Protected branch update failed— that is the guard working, not a broken setup. Don't retry with--force; force pushes to main are blocked too. -
A community PR is never merged here. This procedure covers Sean's own work. Merging someone else's PR requires his explicit per-PR go-ahead — see the
review-prskill's hard rule. -
If CI goes red after a merge, fix forward immediately — don't leave main red.
Why the extra steps
Eight commits landed on main in one session on 2026-07-26 while
enforce_admins was off. Every push printed Bypassed rule violations and
went through anyway; CI ran AFTER each one rather than gating it. All eight
happened to be green, so nothing broke — but a red commit would have landed
exactly the same way. CI takes ~30 seconds. The branch-and-PR round trip costs
about a minute and makes a broken main impossible rather than unlikely.
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.
- 10d ago First seen · 53 lines · 46 tokens per session scan A cde3f042bf60
ship is a skill published in the GitHub repository ShenSeanChen/waku-agent (1,704 stars, last pushed 12d ago), licensed MIT. It adds 46 tokens to every session and 668 once invoked, about $0.0002 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 2026-08-30.
Other skills, from other repositories
github-automation
GitHub workflow automation, PR management, issue tracking, and code review coordination. Integrates with GitHub Actions and repository management. Use when: PR creation, code review, issue management, release automation, workflow setup. Skip when: local-only changes, non-GitHub repositories.
github-repo-management
Clone, create, fork, configure, and manage GitHub repositories. Manage remotes, secrets, releases, and workflows. Works with gh CLI or falls back to git + GitHub REST API via curl.
public-plugin-packaging
Use when packaging an Agentlas agent repo for public GitHub release, Codex plugin submission, Claude adapter distribution, or one-line terminal installation.
git-workflow
Git operations: commits, branches, PRs, and conflict resolution.
graincrawl
Maintain, verify, and release graincrawl, the local-first Granola archive CLI, including SQLite archive behavior, read-only Granola source boundaries, Homebrew tap packaging, and crawlkit-powered TUI/snapshot surfaces.
crawlkit
Maintain and release the crawlkit Go library, preserving downstream compatibility for gitcrawl, slacrawl, discrawl, and notcrawl.