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/taoidle/plan-cascade/claude-mdgit clone --depth 1 https://github.com/Taoidle/plan-cascadeWhat 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 | $0.01231 | $0.01231 |
| Opus 5 | $0.00616 | $0.00616 |
| Sonnet 5 | $0.00246 | $0.00246 |
| Haiku 4.5 | $0.00123 | $0.00123 |
Grade A, and why
plan-cascade CLAUDE.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 3d 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 — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
Plan Cascade is an AI-powered cascading development framework that decomposes complex projects into parallel executable tasks with multi-agent collaboration. It operates as a Claude Code Plugin (stable), MCP Server (stable), Standalone CLI (in development), and Desktop App (in development).
Build & Development Commands
Requires Python 3.10+ and uv
# Run tests (uv auto-creates venv and installs deps)
uv run pytest tests/ # All tests
uv run pytest tests/test_prd_generator.py # Single test file
uv run pytest tests/ -k "test_name" # Run specific test by name
# Linting
uv run ruff check src/
# Type checking
uv run mypy src/
# Initialize external framework skills
git submodule update --init --recursive
# Manual install (alternative to uv)
pip install -e ".[all]"
Entry points: plan-cascade (CLI) and plan-cascade-mcp (MCP server)
Architecture
Three-Tier Execution Model
Mega Plan (project-level) → mega-plan.json + design_doc.json
↓
Hybrid Ralph (feature-level) → prd.json + design_doc.json
↓
Story Execution (task-level) → Code changes
Key Directories
src/plan_cascade/core/- Orchestration engines (PRD generation, parallel execution, quality gates, retry management)src/plan_cascade/backends/- Agent abstraction layer with two implementations:claude_code.py- ClaudeCodeBackend: subprocess-based, no API key neededbuiltin.py- BuiltinBackend: direct LLM API with ReAct loop
src/plan_cascade/state/- Thread-safe state management with platform-specific file lockingsrc/plan_cascade/llm/- LLM provider abstraction (Anthropic, OpenAI, DeepSeek, Ollama)src/plan_cascade/tools/- ReAct tool implementations (file_tools, search_tools, shell_tools)src/plan_cascade/settings/- Configuration storage, validation, and migrationcommands/- Plugin command definitions (markdown files with step-by-step AI instructions)skills/- Plugin skills:hybrid-ralph/,mega-plan/,planning-with-files/mcp_server/- FastMCP server implementationexternal-skills/- Git submodules with framework-specific best practices (React, Vue, Rust)
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.
- 3d ago First seen · 115 lines · 1,231 tokens per session scan A 7c56e174a273
plan-cascade CLAUDE.md is an instructions file published in the GitHub repository Taoidle/plan-cascade (131 stars, last pushed 5mo ago), licensed MIT. It adds 1,231 tokens to every session, about $0.0062 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
Polaris AGENTS.md
Instructions for sponge-b0b/Polaris, covering agents.md, purpose and authority, domain vocabulary, coding conduct and secrets.
llm-council CLAUDE.md
Instructions for amiable-dev/llm-council, covering claude.md — technical notes for llm council, project overview, running & developing, architecture: the l1→l4 layer model (adr-024) and module map (src/llmcouncil/).
brain-in-the-fish CLAUDE.md
Instructions for fabio-rovai/brain-in-the-fish, covering brain in the fish, relationship to open-ontologies, module overview, pipeline and mcp tools.
dos-kernel AGENTS.md
Instructions for anthony-chaudhary/dos-kernel, covering what dos is (the 30-second version), run it yourself first — the 60-second proof, when the user asks you about dos (the consumer moves), build, test, and check your work and the layout in one screen.
dos-kernel GEMINI.md
Instructions for anthony-chaudhary/dos-kernel, covering dos — the part that doesn't believe the agents, the rule, why this matters and setup.
hcifootprint CLAUDE.md
Instructions for footprintjs/hcifootprint, covering hcifootprint — feature-work map, before you design it: it may already exist, module map and gates.