mirror of https://github.com/msitarzewski/agency-agents.git synced 2026-09-22 01:15:54 +03:00

Files

T

Hotragn Pettugani d3a3f573e3 fix(aider): CONVENTIONS.md is a roster index, not 3.8 million characters of agents (#871 )

Aider loads a conventions file into context and keeps it there for the whole
session — that is what the file is for, and the docs say to load it with
--read so prompt caching can hold it. This integration concatenated every
agent body into it. At 279 agents that is 3,816,372 characters, roughly a
million tokens. No model takes that. Anyone who ran

    ./scripts/install.sh --tool aider

got a CONVENTIONS.md that either blows the context window on the first turn
or bills for a million tokens trying.

CONVENTIONS.md is now the roster index it was described as: one entry per
agent with the name, the description, the division, and the path to the
agent file. 96,823 characters, down from 3.8 million. The header explains
how to pull a single agent's full instructions into the session:

    /read-only /path/to/agency-agents/engineering/engineering-frontend-developer.md

Naming an agent in a prompt still works the way it did — the description is
what the model needed for that, and it is still there.

test-convert-outputs.sh now holds the index to being an index: it fails if
CONVENTIONS.md grows past 250,000 characters, if it does not list exactly
one path per roster agent, or if any path it prints does not resolve. The
existing round-trip check on the accumulated file still covers the names and
descriptions.

2026-09-20 18:35:10 -05:00

6.5 KiB

Raw Blame History

🔌 Integrations

This directory contains The Agency integrations and converted formats for supported agentic coding tools.

Supported Tools

Claude Code — .md agents, use the repo directly
GitHub Copilot — .md agents, use the repo directly
Antigravity — SKILL.md per agent in antigravity/
Gemini CLI — .md agent files in gemini-cli/agents/
OpenCode — .md agent files in opencode/
OpenClaw — SOUL.md + AGENTS.md + IDENTITY.md workspaces
Cursor — .mdc rule files in cursor/
Aider — CONVENTIONS.md roster index in aider/
Windsurf — .windsurfrules in windsurf/
Kimi Code — YAML agent specs in kimi/
Qwen Code — project-scoped .md SubAgents in .qwen/agents/
Codex — .toml custom agents in codex/
Mistral Vibe — .toml agents + prompt files generated in vibe/
Osaurus -- SKILL.md skills generated in osaurus/
Hermes -- lazy-router plugin generated in hermes/

Quick Install

# Install for all detected tools automatically
./scripts/install.sh

# Install a specific home-scoped tool
./scripts/install.sh --tool antigravity
./scripts/install.sh --tool copilot
./scripts/install.sh --tool openclaw
./scripts/install.sh --tool claude-code
./scripts/install.sh --tool codex
./scripts/install.sh --tool osaurus
./scripts/install.sh --tool hermes

# Gemini CLI needs generated integration files on a fresh clone
./scripts/convert.sh --tool gemini-cli
./scripts/install.sh --tool gemini-cli

# Qwen Code also needs generated SubAgent files on a fresh clone
./scripts/convert.sh --tool qwen
./scripts/install.sh --tool qwen

If you install OpenClaw and the gateway is already running, restart it after installation:

openclaw gateway restart

For project-scoped tools such as OpenCode, Cursor, Aider, Windsurf, and Qwen Code, run the installer from your target project root as shown in the tool-specific sections below.

Regenerating Integration Files

If you add or modify agents, regenerate all integration files:

./scripts/convert.sh

Claude Code

The Agency was originally designed for Claude Code. Agents work natively without conversion.

cp -r <category>/*.md ~/.claude/agents/
# or install everything at once:
./scripts/install.sh --tool claude-code

See claude-code/README.md for details.

GitHub Copilot

The Agency also works natively with GitHub Copilot. Agents can be copied directly into ~/.github/agents/ and ~/.copilot/agents/ without conversion.

./scripts/install.sh --tool copilot

See github-copilot/README.md for details.

Antigravity

Skills are installed to ~/.gemini/config/skills/. Each agent becomes a separate skill prefixed with agency- to avoid naming conflicts.

./scripts/install.sh --tool antigravity

See antigravity/README.md for details.

Gemini CLI

Agents are packaged as Gemini CLI subagents. Subagents are installed to ~/.gemini/agents/. Because the agent files are generated artifacts, run ./scripts/convert.sh --tool gemini-cli before installing from a fresh clone.

./scripts/convert.sh --tool gemini-cli
./scripts/install.sh --tool gemini-cli

See gemini-cli/README.md for details.

OpenCode

Each agent becomes a project-scoped .md file in .opencode/agents/.

cd /your/project && /path/to/agency-agents/scripts/install.sh --tool opencode

See opencode/README.md for details.

OpenClaw

Each agent becomes an OpenClaw workspace containing SOUL.md, AGENTS.md, and IDENTITY.md.

Before installing, generate the OpenClaw workspaces:

./scripts/convert.sh --tool openclaw

Then install them:

./scripts/install.sh --tool openclaw

See openclaw/README.md for details.

Cursor

Each agent becomes a .mdc rule file. Rules are project-scoped — run the installer from your project root.

cd /your/project && /path/to/agency-agents/scripts/install.sh --tool cursor

See cursor/README.md for details.

Aider

CONVENTIONS.md is the roster index — name, description, and the path to each agent's full instructions. Aider keeps a conventions file in context for the whole session, so the file lists the agents instead of inlining a million tokens of them.

cd /your/project && /path/to/agency-agents/scripts/install.sh --tool aider

See aider/README.md for details.

Windsurf

All agents are consolidated into a single .windsurfrules file for your project root.

cd /your/project && /path/to/agency-agents/scripts/install.sh --tool windsurf

See windsurf/README.md for details.

Kimi Code

Each agent is converted to a Kimi Code CLI agent specification (YAML format with separate system prompt files). Agents are installed to ~/.config/kimi/agents/.

Because the Kimi agent files are generated from the source Markdown, run ./scripts/convert.sh --tool kimi before installing from a fresh clone.

./scripts/convert.sh --tool kimi
./scripts/install.sh --tool kimi

Usage

After installation, use an agent with the --agent-file flag:

kimi --agent-file ~/.config/kimi/agents/frontend-developer/agent.yaml

Or in a specific project:

cd /your/project
kimi --agent-file ~/.config/kimi/agents/frontend-developer/agent.yaml \
     --work-dir /your/project

See kimi/README.md for details.

Qwen Code

Each agent becomes a project-scoped .md SubAgent file in .qwen/agents/.

From a fresh clone, generate the Qwen files first:

./scripts/convert.sh --tool qwen

Then install them from your project root:

cd /your/project && /path/to/agency-agents/scripts/install.sh --tool qwen

See qwen/README.md for details.

Codex

Each agent is converted into a standalone Codex custom agent TOML file and installed to ~/.codex/agents/.

Because Codex uses generated TOML files rather than the source Markdown directly, run the converter before installing from a fresh clone:

./scripts/convert.sh --tool codex
./scripts/install.sh --tool codex

See codex/README.md for details.

6.5 KiB Raw Blame History