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 agents/endogenai/dogma/agent-fleet-design-patternsgit clone --depth 1 https://github.com/EndogenAI/dogmaWrote 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/agents/endogenai/dogma/agent-fleet-design-patterns)<a href="https://agentmods.dev/agents/endogenai/dogma/agent-fleet-design-patterns"><img src="https://agentmods.dev/badge/agents/endogenai/dogma/agent-fleet-design-patterns.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 5.1 | $0.00000 | $0.13164 |
| Opus 5 | $0.00000 | $0.06582 |
| Sonnet 5 | $0.00000 | $0.02633 |
| Haiku 4.5 | $0.00000 | $0.01316 |
Grade A, and why
agent-fleet-design-patterns 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 — 476 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Agent Fleet Design Patterns
Status: Final Research Question: What are the best design patterns for hierarchical agent fleets? How should executives, sub-agents, and specialist agents be structured? Date: 2026-03-06
1. Executive Summary
Across fourteen sources spanning production engineering retrospectives, peer-reviewed research, normative specifications, and practitioner cookbooks, a coherent and mutually reinforcing picture of agent fleet design has emerged. Three design hypotheses were submitted for validation; all three required refinement, but two are confirmed directionally and one required a significant reframe.
The most important single finding is the Compression-on-Ascent / Focus-on-Descent reframe of the original Prompt Enrichment Chain hypothesis. Context does not enrich as it cascades through a delegation hierarchy — it contracts. Lead agents dispatch precise, narrow task briefs downward; subagents explore extensively and then compress findings into dense 1,000–2,000 token handoffs on ascent. Anthropic's production multi-agent research system produced a 90.2% improvement over single-agent baselines not by enriching prompts through the chain, but by multiplying total token budget across parallel isolated context windows. This reframe has direct implications for every agent file in .github/agents/: outbound handoff briefs should be minimal and focused; inbound results should be aggressively compressed.
The second key finding is the confirmation and specification of phase-gated self-loops. Every high-performing agent architecture in the corpus — from ReAct's Thought/Action/Observation cycle, to AIGNE's Constructor/Updater/Evaluator pipeline, to Claude Code's plan-approval gate and hook-based completion enforcement, to Anthropic's three-layer review gate (subagent interleaved thinking → lead re-evaluation → CitationAgent post-processing) — implements explicit evaluation checkpoints before output is accepted and before context advances. The EndogenAI handoff model currently lacks these gates in executable form; they exist as instructions but not as enforced scripts.
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 · 476 lines · 0 tokens per session scan A f825c8884e5f
agent-fleet-design-patterns is an agent published in the GitHub repository EndogenAI/dogma (2 stars, last pushed 12d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 13,164 tokens. 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 2026-09-03.
Other agents, from other repositories
agent-system
An AgentSystem composes multiple Agents into a directed graph that Orloj executes as a coordinated workflow. The graph defines how messages flow between agents during task execution.
acx-implementer
AgentCortex /implement phase executor. Use when delegating implementation work that must follow agentic-os gates, evidence requirements, and skill injection.
acx-tester
AgentCortex /test phase executor. Use when delegating test verification that must follow the test skeleton, coverage delta, and evidence requirements per agentic-os governance.
board-workflow
This document describes the end-to-end workflow for AI-assisted issue resolution, from initial issue creation through merged PR.
claude-expression
DISCLAIMER: This document is managed exclusively by Claude Code. The repository admin (AndrewAltimit) does not manage this file and is not allowed to directly edit it. Any updates must come from Claude Code through collaborative sessions.
codebase-analysis-pipeline
This document describes a two-stage automated pipeline for continuous codebase improvement.