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 instructions/nyrok/flompt/claude-mdgit clone --depth 1 https://github.com/Nyrok/flomptWrote 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/instructions/nyrok/flompt/claude-md)<a href="https://agentmods.dev/instructions/nyrok/flompt/claude-md"><img src="https://agentmods.dev/badge/instructions/nyrok/flompt/claude-md.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 | $0.06460 | $0.06460 |
| Opus 5 | $0.03230 | $0.03230 |
| Sonnet 5 | $0.01292 | $0.01292 |
| Haiku 4.5 | $0.00646 | $0.00646 |
Grade E, and why
flompt CLAUDE.md scanned grade E with 4 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 4d 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.
Sends data to an external URLmediumData exfiltration
A POST to an outside endpoint may be telemetry or may be exfiltration; either way the mod talks to somewhere, and you should know where.
| curl -s -X POST "http://localhost:2019/load" -H "Content-Type: application/json" -d @- Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
- **New routers won't load** until the backend process is fully restarted — verify with `curl -s http://localhost:8000/openapi.json | python3 -c "import sys,json; [print(p) for p in json.load(sys.stdin)['paths']]"` Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
cd /projects/flompt/blog && rm -rf .next && npm run build Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
| curl -s -X POST "http://localhost:2019/load" -H "Content-Type: application/json" -d @- How it starts
The opening of the file, as written. The whole thing — 411 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — flompt
⚠️ Mandatory Rules
- NEVER commit or push
_uploads/directories — they exist at the root, inblog/, and inextension/. These folders contain raw uploaded files and must stay local only. They are in.gitignore; never override or bypass this rule.
Project Identity
- flompt = flow + prompt — Visual AI Prompt Builder
- URL : https://flompt.dev
- Repo : https://github.com/Nyrok/flompt (monorepo)
- Git email : [email protected]
Stack
- App : React 18 + TypeScript + React Flow v11 + Zustand + Vite (SPA in
/app) - Blog : Next.js 16 + Tailwind CSS (static export in
/blog, bilingual FR/EN) - Landing : Static HTML (in
/landing) - Backend : FastAPI + Uvicorn (Python 3.12, port 8000)
- Reverse Proxy : Caddy (auto-TLS Let's Encrypt, port 443)
- AI : Anthropic Claude (pluggable, via httpx) + Groq (Llama Guard 4 prompt safety — currently DISABLED via
PROMPT_GUARD_ENABLED=falseinbackend/.env) - MCP : FastMCP server (
backend/app/mcp_server.py) exposes decompose/compile tools — registered on Glama.ai - Analytics : PostHog (EU region) — autocapture, session replay, heatmaps, error tracking
- i18n : 10 languages (EN FR ES DE PT JA TR ZH AR RU) via LocaleContext + JSON files. Locale priority: URL path (
/app/fr) → localStorage → default'en' - SEO : Static locale pages generated post-build (
app/scripts/generate-locale-pages.js). Each/app/[locale]serves a dedicated HTML with localized title, description, canonical and hreflang×10
Deployment Architecture
flompt.dev/
├── / → landing/index.html (static, catch-all)
├── /app* → app/dist/ (Vite SPA, handle_path strips /app)
├── /blog* → blog/out/ (Next.js static export, handle_path strips /blog)
├── /api/* → FastAPI :8000 (reverse_proxy)
└── /health → FastAPI :8000 (reverse_proxy)
Monorepo Structure
/projects/flompt/
├── app/ # Vite React SPA (prompt builder)
│ ├── src/ # React components, styles, store
│ ├── dist/ # Production build (gitignored)
│ └── index.html # Entry point (base: /app)
├── blog/ # Next.js blog (static export)
│ ├── src/ # Pages, components, i18n
│ ├── content/ # Markdown posts (fr/ + en/) — slugs MUST match between locales
│ ├── out/ # Static export (gitignored)
│ └── next.config.ts # basePath: /blog, output: export
├── landing/ # Static landing page
│ └── index.html
├── backend/ # FastAPI backend
│ ├── app/
│ └── .venv/
├── extension/ # Browser extension (Chrome + Firefox)
│ ├── Makefile # make = both; make chrome; make firefox
│ └── dist/ # Built zips (gitignored)
├── docs/ # Documentation (block-types, chrome-extension, how-it-works, claude-code)
├── marketing/ # Marketing strategy, post drafts, activity logs
├── deploy.sh # Full redeploy script (builds app+blog, restarts backend, reloads Caddy, health checks)
├── docker-compose.yml # Local development (backend + frontend services)
├── supervisord.conf # Production process manager (backend + Caddy)
├── watchdog.sh / keepalive.sh # Supervisor health monitoring
├── Caddyfile # Reverse proxy config
├── caddy # Caddy binary (gitignored, 50MB)
└── CLAUDE.md # This file
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.
- 4d ago First seen · 411 lines · 6,460 tokens per session scan E fc975bca080b
flompt CLAUDE.md is an instructions file published in the GitHub repository Nyrok/flompt (149 stars, last pushed 5mo ago), licensed MIT. It adds 6,460 tokens to every session, about $0.0323 per session on Opus 5. A static security scan graded it E with 4 findings (sends data to an external url, downloads and executes remote code, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
skillkit CLAUDE.md
Instructions for maxvaega/skillkit, covering skillkit, development approach, current focus (v0.4), key architectural decisions and core architecture (v0.1).
claude-prompt AGENTS.md
Instructions for danielkremen818/claude-prompt, covering agents.md, purpose, structure map, invariants (keep green) and conventions.
Promptheus CLAUDE.md
Instructions for abhichandra21/Promptheus, covering claude.md, development commands, installation and setup, install in development mode and or use the traditional setup.py approach.
Promptheus AGENTS.md
Instructions for abhichandra21/Promptheus, covering repository guidelines, project structure & module organization, build, test, and development commands, quiet mode for scripting (clean stdout/stderr separation) and different output formats.
Promptheus GEMINI.md
Instructions for abhichandra21/Promptheus, covering promptheus project context, 1. project overview, core features, architecture and 2. building and running.
claude-prompt CLAUDE.md
Instructions for danielkremen818/claude-prompt: This file guides AI agents (and humans) working on the claude-prompt repository itself.