What is hive AGENTS.md?

A project guide for Hive, a server that gives AI coding assistants access to an Obsidian note vault and can delegate work to local or low-cost workers. It also documents the repository’s architecture and operating rules.

Use it when working on the Hive repository, especially before changing code, following its development process, or troubleshooting its orchestration setup.

It tells an assistant which instructions and documentation to read first, where project decisions and task status are stored, and how worker routing works.

hive AGENTS.md — Instructions file by mlorentedev

Install

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.

agentmods

npx agentmods add instructions/mlorentedev/hive/agents-md

Clone the repo

git clone --depth 1 https://github.com/mlorentedev/hive

Made for: Codex, OpenCode.

Per session 3,040 This file is loaded in full into every session.

When invoked 3,040 The same file — it is already loaded in full.

Security scan A 1 finding. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

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	$0.03040	$0.03040
Opus 5	$0.01520	$0.01520
Sonnet 5	$0.00608	$0.00608
Haiku 4.5	$0.00304	$0.00304

Measured 2d ago against content hash 2c3e3fdca43c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

hive AGENTS.md scanned grade A with 1 finding 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.

Runs shell commandslowCapability

Expected in a hook, worth knowing in a rule or an instructions file.

- **All `subprocess.run` calls MUST catch broad `Exception`**, not just `CalledProcessError` — git invocations on Windows raise things like `FileNotFoundError` that would otherwise crash a write tool.

AGENTS.md · 132 lines

How it starts

The opening of the file, as written. The whole thing — 132 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md

Instructions for AI coding agents (Claude Code, OpenCode, Copilot, Cursor, Codex, Antigravity) operating in this repo.

Behavioural SSOT lives in dotfiles: $DOTFILES_REPO_DIR/AGENTS.md (the dotfiles repo root, resolved per-machine via the path cascade: explicit env → ~/.config/dotfiles/machine.json → env-contract.json default; ADR-025). Read it FIRST — Identity, Standing Orders, Decision Hierarchy, Model Selection, Neural Hive protocol, MCP usage rules, Spec-Driven Development gate, and operational rules. This file adds ONLY what is specific to the Hive repo. The Claude-specific tooling overlay is .claude/CLAUDE.md.

What this repo is

Hive — vault-native AI orchestration: a unified MCP server giving AI coding assistants on-demand access to an Obsidian vault plus delegation to local/cheap workers.

Build/operate docs live in docs/ (docs-as-code): docs/adr/ (architecture decisions + sequence-diagrams.md), docs/runbooks/, docs/troubleshooting/, docs/lessons.md. Per-feature specs live in specs/. Task state lives in the bitácora GitHub Project (filter by Repo = hive), not here. The cross-project brain and AI memory live in the maintainer's vault.

Big-picture architecture

Hive is an MCP server (stdio transport, FastMCP framework) with three responsibilities:

Vault tools — query, search, list, write, patch markdown files in an Obsidian vault. All writes auto-commit to git (best-effort; git failure never crashes the server) and validate YAML frontmatter.
Session tools — session_briefing assembles tasks + lessons + git log + health in one call so an AI client gets ~50 lines of context instead of ~800.
Worker tools — delegate_task and capture_lesson each make one attempt against one OpenAI-compatible worker. No internal fallback: the caller owns the chain, and failures are classified as pool unavailable vs task failed so the caller can tell a retryable one from a real one.

Read the full file on GitHub · 132 lines

Changes

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 · 132 lines · 3,040 tokens per session scan A 2c3e3fdca43c

Subscribe to this mod's changes

hive AGENTS.md is an instructions file published in the GitHub repository mlorentedev/hive (10 stars, last pushed 7d ago), licensed MIT. It adds 3,040 tokens to every session, about $0.0152 per session on Opus 5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

ai claude-code context-management developer-tools mcp obsidian

hive AGENTS.md

Getting it into your agent

What it costs to keep this loaded

Grade A, and why

Runs shell commandslowCapability

How it starts

AGENTS.md

What this repo is

Big-picture architecture

What this file has done since we first saw it

Other instructions, from other repositories

plannotator AGENTS.md

textbook-to-note CLAUDE.md

obsidian-llm-hub AGENTS.md

kObsidian AGENTS.md

obsidian-pkm-plugin CLAUDE.md

llm-wiki copilot-instructions.md