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/icohangar-ops/agent-conductor/agents-mdgit clone --depth 1 https://github.com/icohangar-ops/agent-conductorWrote 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/icohangar-ops/agent-conductor/agents-md)<a href="https://agentmods.dev/instructions/icohangar-ops/agent-conductor/agents-md"><img src="https://agentmods.dev/badge/instructions/icohangar-ops/agent-conductor/agents-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 | $0.00606 | $0.00606 |
| Opus 5 | $0.00303 | $0.00303 |
| Sonnet 5 | $0.00121 | $0.00121 |
| Haiku 4.5 | $0.00061 | $0.00061 |
Grade A, and why
agent-conductor 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 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.
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 — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Agent Conductor
Guidance for AI coding agents working in this repository. (Yes — Conductor's
own contract compiles with Conductor: contract_load on this file.)
Mission
Agent Conductor turns AGENTS.md operating manuals and SKILL.md skills into an active orchestration layer over MCP, with the Consensus Hardening Protocol gating high-stakes decisions. Agents should keep the parser lossless, the dependency surface near zero, and the MCP stdout stream protocol-clean.
Architecture
| Layer | Role | Do | Don't |
|---|---|---|---|
src/contract/ |
Contract compilation | Extract recognized AGENTS.md patterns; preserve unknown sections | Drop content the parser doesn't understand |
src/skills/ |
Skill discovery | Frontmatter-only metadata on discovery; bodies on demand | Load bodies eagerly |
src/engine/ + engine/bridge.py |
Decision engine boundary | Speak newline-delimited JSON over child stdio | Print logs to stdout on either side |
PyPI consensus-hardening-protocol |
Published CHP engine (import chp) |
Pin in engine/requirements.txt; fix upstream |
Re-vendor a private copy |
test/ |
Regression guard | Test against the real example fixture | Mock the Python engine in engine tests |
Engineering rules
Non-negotiables
- stdout is sacred — MCP transport owns process stdout; all logging goes to stderr on both sides of the bridge.
- Zero new runtime dependencies — markdown/frontmatter parsing stays hand-rolled; only the MCP SDK and zod are allowed.
- Lossless compilation — unrecognized AGENTS.md sections must survive into
contract.sections. - Erasable TypeScript only — source must run under Node's type stripping (no enums, no parameter properties);
tsconfigenforceserasableSyntaxOnly. - CHP via PyPI — decision logic comes from
consensus-hardening-protocol(pip install -r engine/requirements.txt). Do not re-vendor underengine/. - Engine needs Python 3.10+ — install the published package before running bridge / MCP decision tools.
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 · 56 lines · 606 tokens per session scan A 483e6df9771a
agent-conductor AGENTS.md is an instructions file published in the GitHub repository icohangar-ops/agent-conductor (0 stars, last pushed 13d ago), licensed MIT. It adds 606 tokens to every session, about $0.0030 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
enquire-mcp AGENTS.md
AGENTS.md instructions for oomkapwn/enquire-mcp, covering agents.md, tl;dr, architecture (5-minute orientation), conventions and typescript.
repo-context-mcp AGENTS.md
Instructions for nduc99911/repo-context-mcp, covering agents.md — repo-context-mcp, project, layout, commands and conventions.
toon-memory AGENTS.md
Instructions for LuiggiVal08/toon-memory, covering agents.md, build & verify, architecture, smarter memory pattern and visual memory pattern.
moo-tasks AGENTS.md
AGENTS.md instructions for shekarsiri/moo-tasks, covering 🐮 agent guidelines & protocol (moo tasks), ⚡ 1-call fast-path for rapid coding (recommended), 🎯 full mandatory 6-step workflow protocol, 1. goal anchoring (prevent scope drift) and 2. task planning & full specifications.
mcp-yandex-audience CLAUDE.md
Claude Code instructions for A1-x-Tech/mcp-yandex-audience, covering claude.md — mcp-yandex-audience, commands, architecture, conventions (do not break) and adding a tool.
mcp-yandex-merchants CLAUDE.md
Claude Code instructions for A1-x-Tech/mcp-yandex-merchants, covering claude.md — mcp-yandex-merchants, commands, architecture, conventions (do not break) and adding a tool.