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 skills add tzavislan/kumihimo --skill kumihimo-retrogit clone --depth 1 https://github.com/tzavislan/kumihimoWrote 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/tzavislan/kumihimo/kumihimo-retro)<a href="https://agentmods.dev/skills/tzavislan/kumihimo/kumihimo-retro"><img src="https://agentmods.dev/badge/skills/tzavislan/kumihimo/kumihimo-retro/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/tzavislan/kumihimo/kumihimo-retro"><img src="https://agentmods.dev/badge/skills/tzavislan/kumihimo/kumihimo-retro.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00069 | $0.00686 |
| Opus 5 | $0.00034 | $0.00343 |
| Sonnet 5 | $0.00014 | $0.00137 |
| Haiku 4.5 | $0.00007 | $0.00069 |
Grade A, and why
kumihimo-retro 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 — 54 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Kumihimo retro — train the tooling
The iteration skill only stays good if what the loop learns gets written back
into it. Yorishiro's build-iteration skill accreted its best rules (the
blocking-rounds cap, the git add -A incident) from real failures; this skill
makes that folding a scheduled obligation instead of an accident.
When
At each milestone close, or when loop.json's iteration is 10+ past
last_retro_iteration — whichever comes first.
The pass
- Read
build/state/journal.mdfrom the entry afterlast_retro_iterationto now, plus anyneeds-thomasitems in the queue. - Extract durable lessons only. A lesson is durable when it would change what the next iteration does: a mistake made twice, a verification step that caught something (or failed to), a convention that turned out ambiguous, a command that should be in the battery. One-off trivia stays in the journal.
- Fold each lesson into the right home, smallest diff that carries it:
- how to run an iteration →
.claude/skills/kumihimo-iteration/SKILL.md - how to manage the project (status, queue, milestones, releases) →
.claude/skills/kumihimo-manage/SKILL.md - a repo rule or invariant →
CLAUDE.md - a coding/documentation rule →
CONVENTIONS.md(and, if enforceable, queue a linter check for it — unenforced rules rot) - background Thomas should have across sessions → Claude's project memory
- how to run an iteration →
- Prune guidance that events proved wrong or stale. A skill that only grows becomes noise; deleting a dead rule is as valuable as adding a live one.
- Record: append a
retroentry to the journal naming each lesson and where it landed (or why it was dropped), append a dated line to the Training log in.claude/skills/kumihimo-manage/SKILL.md(the unbroken log is the proof the loop is alive), setlast_retro_iterationinloop.json, and commit everything asretro: <n> lessons through iteration <i>.
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 · 54 lines · 69 tokens per session scan A 28730dcb08d8
kumihimo-retro is a skill published in the GitHub repository tzavislan/kumihimo (0 stars, last pushed 6d ago), licensed MIT. It adds 69 tokens to every session and 686 once invoked, about $0.0003 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
worktree-execution
Creates, pins, and cleans isolated Git worktree pairs for comparison trials. Use only when a Rashomon prompt or skill evaluation needs paired execution environments or orphan recovery.
lean-commit
Ultra-compressed commit message generator. Cuts noise from commit messages while preserving intent and reasoning. Conventional Commits format. Subject ≤50 chars, body only when "why" isn't obvious. Use when user says "write a commit", "commit message", "generate commit", "/commit", or invokes…
release-checklist
Minimal release checklist used as the control variant.
lfe-archivist
Act as the Archivist for an LFE-compliant project. Use when a change is approved and needs documentation sync, changelog updates, or pipeline status cleanup.
git-helper
Generate git commit messages and help with git workflows.
agent-teacher
A teaching aid that explains technical ideas with a small runnable code example and a guided walkthrough. It is meant for learning how something works, not fixing existing code.