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/andreidrang/python3-capsolver/agents-mdgit clone --depth 1 https://github.com/AndreiDrang/python3-capsolverWhat 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 | $0.00952 | $0.00952 |
| Opus 5 | $0.00476 | $0.00476 |
| Sonnet 5 | $0.00190 | $0.00190 |
| Haiku 4.5 | $0.00095 | $0.00095 |
Grade A, and why
python3-capsolver AGENTS.md 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 — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Repository overview
Python 3.8+ client library for the Capsolver captcha-solving API. Single-package src/ layout, dual sync (requests) / async (aiohttp) execution, msgspec for serialization, tenacity for async retries.
Where to work
./
├── src/python3_capsolver/ # Main library
│ ├── core/ # Base classes, instruments, serializers, enums
│ ├── control.py # Raw API: get_balance, create_task, get_task_result
│ ├── recaptcha.py # ReCaptcha V2/V3/Enterprise
│ ├── cloudflare.py # Cloudflare Turnstile/Challenge
│ └── *.py # Other captcha services (see package AGENTS.md)
├── tests/ # Pytest suite mirroring source structure
│ ├── conftest.py # BaseTest class, fixtures, rate-limiting delays
│ └── test_*.py # One file per service + test_core.py + test_instrument.py
├── docs/ # Sphinx documentation (make doc)
├── ARCHITECTURE.md # Layered architecture, data flow, invariants
├── pyproject.toml # Build, deps, black/isort/pytest config
└── Makefile # make tests, make refactor, make build, make doc
Architecture and boundaries
Four layers with strict dependency direction (top → bottom only):
- Service layer (
src/python3_capsolver/*.py) — thin wrappers inheritingCaptchaParams, zero HTTP logic - Base layer (
core/base.py) —CaptchaParamsmerges payloads, delegates to instruments - Instrument layer (
core/*_instrument.py) —SIOCaptchaInstrument(sync) andAIOCaptchaInstrument(async) handle all HTTP, retries, and polling - Support layer (
core/serializer.py,core/enum.py,core/const.py) —msgspec.Structclasses, enums, constants
Forbidden: service files importing requests/aiohttp directly; support layer depending on upper layers.
Change rules
- Every new captcha service must inherit
CaptchaParams, providecaptcha_handler()+aio_captcha_handler(), and register its type inCaptchaTypeEnm - All serialization must use
msgspec.Struct— never rawjsonmodule - Dual sync/async is mandatory for any new instrument or service
- Service classes are thin: only
__init__with captcha-type-specific params; all HTTP goes through instruments - Context manager support (
with/async with) is required on all service classes
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 · 70 lines · 952 tokens per session scan A 19236e170c00
python3-capsolver AGENTS.md is an instructions file published in the GitHub repository AndreiDrang/python3-capsolver (91 stars, last pushed 24d ago), licensed MIT. It adds 952 tokens to every session, about $0.0048 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 2026-08-30.
Other instructions, from other repositories
terraform-provider-akamai copilot-instructions.md
Copilot instructions for akamai/terraform-provider-akamai, covering ai review guidelines — terraform-provider-akamai, meta information for ai agents, 1. general practices, 2. common coding practices (apply to every change) and 3. changelog rules.
2Captcha-MCP CLAUDE.md
Claude Code instructions for 0-SOFT/2Captcha-MCP, covering claude.md — project rules for twocaptcha-mcp, what this project is, critical principles, repository layout and conventions for adding a new captcha type.
METTLE CLAUDE.md
Claude Code instructions for Creed-Space/METTLE, covering claude.md: mettle, project overview, key structure, bilateral alignment and wiki knowledge base.
cpf-validador CLAUDE.md
Instructions for opastorello/cpf-validador, covering claude.md, comandos, instalar dependências, rodar localmente (fastapi + fastmcp na porta 8000) and docker (porta 8002 conforme docker-compose.yaml).
deepspider CLAUDE.md
Claude Code instructions for ma-pony/deepspider, covering deepspider repository guidance, runtime boundaries, reverse-engineering contract, tool schemas and commands.
Agent-skills CLAUDE.md
Claude Code instructions for Captcha-Sonic/Agent-skills, covering captchasonic plugin for claude code, what this plugin does, mcp tools available, setup and skills.