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/ribeirogab/specwright/shipnpx skills add ribeirogab/specwright --skill shipgit clone --depth 1 https://github.com/ribeirogab/specwrightWhat 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 | $0.00076 | $0.00942 |
| Opus 5 | $0.00038 | $0.00471 |
| Sonnet 5 | $0.00015 | $0.00188 |
| Haiku 4.5 | $0.00008 | $0.00094 |
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 2d 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 — 85 lines — stays where its author put it; the contents beside it link to each section on GitHub.
ship — the ladder, without stops
Run every step from intent to a reviewed branch in one go, asking nothing along the way. This is the same ladder as the individual commands — it does not reimplement any step, it runs them back to back and takes on itself every decision the interactive path would have handed to the maintainer.
Announce at start: "Shipping this end to end — I'll decide open questions and record each one."
The steps
Run these in order, each per its own skill. Follow that skill's instructions exactly; the only difference is that you never stop to ask.
propose— writeproposal.mdwith acceptance criteria. If a change folder already exists for this work, start from it instead of writing a new one.plan— decide the scope, writetasks.mdand, abovelow,design.md, then pass the handoff-readiness validator.implement— work the task list, quality gate, runtime verification.review— review the branch, fix what the findings justify, iterate tolgtm.
Then set status: shipped and the shipped: date in proposal.md, tick the
criteria that were verified, and commit on the change's own branch.
Decide, then record
Every question the interactive ladder would have asked, you answer. Two rules:
Prefer the reversible option. When two designs are defensible, take the one that is cheaper to undo. An autonomous run is not the place to make an expensive bet on the maintainer's behalf.
Record every decision in proposal.md. Under ## Decisions and discoveries,
one bullet each, saying what was chosen, why, and what was rejected:
- **[decision]** Used cursor pagination instead of offset — the ticket did not
specify and the table exceeds a million rows. Rejected: offset, which degrades
with depth.
That section is the entire audit trail of an unattended run. A decision you took and did not write down is indistinguishable from one you never noticed — treat an unrecorded decision as a defect, not as tidiness.
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.
- 2d ago First seen · 85 lines · 76 tokens per session scan A 50b99e2cf970
ship is a skill published in the GitHub repository ribeirogab/specwright (3 stars, last pushed 24d ago), licensed MIT. It adds 76 tokens to every session and 942 once invoked, about $0.0004 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
pushing-commits-to-the-repo
Open and advance a PR — write a current title and body, label it, review before every push, watch CI, and triage every comment. Use whenever you open a PR or push a commit to one.
gsd-audit-fix
Autonomous audit-to-fix pipeline — find issues, classify, fix, test, commit.
commit
Generate git commit messages from staged changes.
commit-generate
Generate bilingual (English + Traditional Chinese) commit message from git changes.
trellis-finish-work
Wrap up the current session: verify quality gate passed, remind user to commit, archive completed tasks, and record session progress to the developer journal. Use when done coding and ready to end the session.
git-github-flow
Use for branch prep, clean commits, PR descriptions, GitHub issue triage, changelogs, release notes, review response, merge readiness, or publishing a branch safely.