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/paultyng/ideate/agents-mdgit clone --depth 1 https://github.com/paultyng/ideateWhat 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 | $0.00641 | $0.00641 |
| Opus 5 | $0.00320 | $0.00320 |
| Sonnet 5 | $0.00128 | $0.00128 |
| Haiku 4.5 | $0.00064 | $0.00064 |
Grade A, and why
ideate 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 — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Cross-agent guidance for any AI agent (Claude Code, future runners, external orchestrators) working on the Ideate codebase. Claude-specific context lives in CLAUDE.md; this file captures principles that apply regardless of which agent is at the keyboard.
Build features as MCP tools first, UI second
When adding a new capability, the default surface is an MCP tool on the orchestrator (root orchestrator) MCP server, not a UI control.
Why:
- The orchestrator is the most-used surface — most workflows already start there. A tool exposed there reaches the user without a context switch.
- Tools compose: the orchestrator can chain
rename_idea,add_resource_by_slug,update_idea_by_sluginto a single operation. UI buttons can't. - Tools are scriptable. A workflow that lives only in the UI can't be triggered from a session, a hook, or another agent.
- Less code: an MCP tool is a handler + a schema. A UI feature adds React state, routing, styling, Playwright tests, and the back-and- forth between frontend and backend.
The default lift is therefore: pick the right MCP surface (per-idea vs orchestrator), wire the tool, write the test. Stop there.
When to add UI
Only when the operation cannot reasonably be driven from a conversation. The clearest case is starting an interactive session — the user has to physically attach to a PTY, so a button is unavoidable. Other cases that have justified UI in the past:
- Live terminal interaction (the xterm canvas is the UI).
- Markdown / diff review surfaces — the human reads a rendered diff and clicks comment widgets that don't translate to text.
- Status indicators that need to stay visible across navigations (footer toggles, session bar chips).
If a feature is a one-shot data mutation ("rename this idea", "drop old reviews", "add a tag") it belongs in MCP. If it's a continuous visual state ("show me the agent's terminal", "render this diff"), it belongs in the UI.
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 · 64 lines · 641 tokens per session scan A 751b10b7823f
ideate AGENTS.md is an instructions file published in the GitHub repository paultyng/ideate (9 stars, last pushed 15d ago), licensed MIT. It adds 641 tokens to every session, about $0.0032 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 instructions, from other repositories
wayland AGENTS.md
AGENTS.md instructions for FerroxLabs/wayland, covering coordination (read every task — multi-agent blackboard), wayland - project guide, code conventions, file & directory structure and naming.
wayland copilot-instructions.md
Copilot instructions for FerroxLabs/wayland, covering ijfw rules, output discipline, memory routing, context discipline and cross-audit.
wayland CLAUDE.md
Claude Code instructions for FerroxLabs/wayland, a project described as: Wayland - The AI Agent That Perceives. Reasons. Acts. Evolves.
Herdforge CLAUDE.md
Instructions for Kampe/Herdforge, covering claude.md - build bootstrap & routing protocol, 1. quick invariants (hard rules), 2. fast commands, build binary and run preflight, unit tests, and build.
psyche-build AGENTS.md
Instructions for OpenCoven/psyche-build, covering psyche build agent guide, repository role, planning and evidence contract, canonical routing and deterministic entrypoints.
Herdforge AGENTS.md
Instructions for Kampe/Herdforge, covering herdforge agentic worker contract, 1. core engineering invariants & rules, 2. core package ownership grid (31 of 109 packages), 3. workflow for agentic tasks and code intelligence: code-review-graph cli.