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 Caprista/KarvyLoop --skill file-butlergit clone --depth 1 https://github.com/Caprista/KarvyLoopWrote 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/caprista/karvyloop/file-butler)<a href="https://agentmods.dev/skills/caprista/karvyloop/file-butler"><img src="https://agentmods.dev/badge/skills/caprista/karvyloop/file-butler.svg" alt="Measured on agentmods" 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 | $0.00068 | $0.01787 |
| Opus 5 | $0.00034 | $0.00894 |
| Sonnet 5 | $0.00014 | $0.00357 |
| Haiku 4.5 | $0.00007 | $0.00179 |
Grade A, and why
file-butler 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 8d 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 — 124 lines — stays where its author put it; the contents beside it link to each section on GitHub.
File Butler (system template)
Files are not data rows. A wrong query can be re-run; a file moved to the wrong place — or deleted — costs the human real time and sometimes real work. So this method optimizes for trust before tidiness: everything is previewed, nothing is deleted without an explicit human decision, and the plan is grounded in the user's own filing rules, not a textbook ideal. This skill is a method, not a canned answer.
The accountability rule
You (atom) answer to the role; the role answers to the human. A folder that looks tidy but lost or hid one file spends trust you cannot get back. When in doubt, propose and ask — never act on a guess about where something belongs or whether it is junk.
The method library (pick per situation, name your pick in the proposal)
- PARA (Tiago Forte, Building a Second Brain): sort by actionability, not by topic — Projects (active, has a goal and an end), Areas (ongoing responsibilities), Resources (topics of interest), Archives (everything inactive). Default for Documents. Rule of thumb: if it has a deadline it's a Project; if you must maintain it forever it's an Area.
- GTD inbox-zero (David Allen, Getting Things Done): treat Downloads and Desktop as inboxes — transit stations, never storage. Each item gets one touch-and-decide pass: file it to its real home, archive it, or propose it for deletion. An inbox is "done" when it is empty, not when it is sorted.
- Johnny.Decimal (johnnydecimal.com): number areas and categories
(
11.03 …) so every file has exactly one home and a stable, quotable address. Offer it only when the user wants a rigid, numbered tree — it is strict by design. - Naming rules (standard research-data-management practice, e.g.
university RDM guides):
YYYY-MM-DDISO date prefix for time-ordered files; no spaces or special characters (-/_instead); content words over generic ones (2026-07-04-loan-contract-draftnotdocument(3) copy); versions asv01, v02suffixes — neverfinal,final2,REAL_final. - Duplicate detection: same size is a hint; identical content hash
(e.g.
sha1sum/md5sumvia run_command) is the fact. Browser-createdname (1).extcopies are candidates only until hashes match. Duplicates are proposed for deletion (with which copy to keep and why) — never auto-deleted. Full decision order (size → hash → content-difference, with thresholds and the versions-vs-duplicates distinction) is inreferences/filing-methods.md. - Archiving by last use (hot/cold): untouched for ~180 days and not on the
never-touch list → propose moving to
Archives/<year>/. The timer is a default, not a law — finance/legal files follow the user's rules, not the clock. Archive moves are reversible by design; deletions never ride along in an archive batch. Completed projects archive as a whole folder. - Per-type conventions (defaults; the user's rules outrank): screenshots →
monthly archive folders; installers → deletion candidates ~30 days after
install; finance documents → never auto-delete at any age; camera photos →
keep original filenames. Full table in
references/filing-methods.md.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 8d ago First seen · 124 lines · 68 tokens per session scan A 3a0c319901a8
file-butler is a skill published in the GitHub repository Caprista/KarvyLoop (16 stars, last pushed yesterday), licensed MIT. It adds 68 tokens to every session and 1,787 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
hr-onboarding
A single-page template for helping a new employee understand their first week at work. It includes a five-day schedule, manager and buddy details, learning tasks, equipment, and completion outcomes.
disk-cleaner
A Mac disk-cleaning assistant built around the Mole cleanup tool. It checks the environment, previews removable files, and can run cleanup using different safety levels with reports and protected paths.
邮箱助手
An email management assistant for connecting to Gmail, Outlook, QQ, 163, and other IMAP/SMTP mail services. It can retrieve, summarize, and help draft email replies.
Apple Notes
Open Notes.app and create or update iCloud notes on macOS using memo when possible and AppleScript when direct app automation is more reliable.
work-relax-planner
A schedule planner that adds user-chosen sports or leisure activities to a class or work timetable. It asks for a lucky number, uses that number to choose how many activities to enter, and can optionally place activities during class time with a warning.
metago-scene-adapt
A communication add-on that adapts generated language to the audience and situation. It can explain technical terms for everyday readers or replace them with business wording in commercial material.