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/benmarte/autoimprove/improve-loopnpx skills add benmarte/autoimprove --skill improve-loopgit clone --depth 1 https://github.com/benmarte/autoimproveWhat 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 | $0.00078 | $0.02495 |
| Opus 5 | $0.00039 | $0.01247 |
| Sonnet 5 | $0.00016 | $0.00499 |
| Haiku 4.5 | $0.00008 | $0.00249 |
Grade C, and why
improve-loop scanned grade C with 1 finding 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
rm -rf .claude/autoimprove/worktrees How it starts
The opening of the file, as written. The whole thing — 275 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AutoImprove Loop Skill
Every experiment runs in an isolated git worktree. The main codebase is never modified during experiments. Only winning changes get squash-merged back.
Main branch ──────────────────────────────────── (never touched mid-session)
│ │
experiment-001 experiment-002
(kept ✅ → merge) (discarded ❌ → deleted)
Pre-flight checks
Before the first iteration, print each check as you run it:
━━━ Pre-flight ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✓ Config found
✓ Git working tree clean
✓ Base commit: abc1234
✓ Worktree directory ready
✓ Baseline score: XX/100
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- Check
.claude/autoimprove/config.mdexists. If not, stop: "Run /autoimprove:setup first." - Check git is available:
git status - Confirm main working tree is clean. If not, stop: "Please commit or stash changes before running autoimprove."
- Record the base commit:
git rev-parse HEAD— all experiments branch from here. - Run the worktree skill's setup step to create
.claude/autoimprove/worktrees/and update.gitignore. - Run the measure skill in the main directory to get the BASELINE score.
- Report: "Baseline: XX/100. All experiments will run in isolated worktrees. Main branch is safe."
Session Header
After pre-flight passes, write a session header to .claude/autoimprove/log.md:
## Session — [ISO 8601 timestamp]
**Planned:** N iterations
**Focus:** "focus string" (or "all improvement areas" if none)
**Baseline:** XX/100
**Base commit:** [full SHA]
**Status:** IN_PROGRESS (0/N completed)
If the log file doesn't exist, create it with the project header first:
# .claude/autoimprove/log.md
> Generated by [autoimprove](https://github.com/benmarte/autoimprove) — Claude Code Plugin
> Project: **[project name]** · Stack: [detected stack] · Started: [date]
---
Then append the session header.
Continue Mode
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 · 275 lines · 0 tokens per session scan C 95116134c027
improve-loop is a skill published in the GitHub repository benmarte/autoimprove (5 stars, last pushed 5mo ago), licensed MIT. It adds 78 tokens to every session and 2,495 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
srt-whiteboard-animation
将 SRT 字幕做成暖米黄纸张底的白板手绘动画：读字幕→输出配图策略→确认后生成统一风格线稿→按叙事语义标注分区→预览台调整→渲染 MP4。编排沿用分区遮罩揭示（annotation.json / sequence / startMs / protectedRegions），但每个区域内的落墨换成 stream 的连续笔迹（骨架/网格 ink→color）。当用户提供 SRT 字幕并要求"字幕做成白板手绘/流式笔迹视频""SRT 生成白板动画""按字幕分镜画手绘"时触发。.
user-research-cookiy
End-to-end user research assistant — qualitative and quantitative. Use this skill whenever the user mentions user research, user interviews, discussion guides, interview guides, research plans, qualitative research, quantitative research, user surveys, survey design, usability studies, participant recruitment…
seo
Deterministic LLM-first SEO audits for websites, blog posts, and GitHub repositories. Use this when the user asks to "perform SEO analysis", "run SEO audit", "analyze SEO", "check technical SEO", "review schema", "Core Web Vitals", "E-E-A-T", "hreflang", "GEO", "AEO", or GitHub repository SEO optimization. For…
playwright-best-practices
Use when writing Playwright tests, fixing flaky tests, debugging failures, implementing Page Object Model, configuring CI/CD, optimizing performance, mocking APIs, handling authentication or OAuth, testing accessibility (axe-core), file uploads/downloads, date/time mocking, WebSockets, geolocation, permissions…
open-map-stack
Use textual agent instructions for GIS and geospatial work: source discovery and provenance, vector/raster/point-cloud pipelines, CRS and metric analysis, spatial SQL, routing and isochrones, QGIS projects, tile generation, and web maps. Use advanced tools and formats such as OSM, Overture, STAC, Sentinel/Landsat…
globalpercent
GlobalPercent — build a global-macro-probability panel for an investment research system. Merges public probability data from prediction markets (Polymarket + Kalshi), classifies every market into macro modules (monetary policy / macro economy / AI / etc.), and shows the whole market's expected-probability state at a…