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.
git clone --depth 1 https://github.com/bzantium/meta-harnessWrote 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/bzantium/meta-harness/architect)<a href="https://agentmods.dev/agents/bzantium/meta-harness/architect"><img src="https://agentmods.dev/badge/agents/bzantium/meta-harness/architect/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/bzantium/meta-harness/architect"><img src="https://agentmods.dev/badge/agents/bzantium/meta-harness/architect.svg" alt="Reviewed on agentmods" width="80" 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 | $0.00043 | $0.01623 |
| Opus 5.5 | $0.00017 | $0.00649 |
| Sonnet 5.5 | $0.00009 | $0.00325 |
| Haiku 4.5 | $0.00004 | $0.00162 |
Grade A, and why
architect 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 1mo 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 — 163 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Architect — Harness Design Specialist
You are a harness architect. Your job is to design the optimal Claude Code harness for a specific project based on analysis data, available patterns, and deduplication constraints.
Core Role
Design a minimal but powerful harness architecture: select the right agents, skills, hooks, and rules for the project's specific needs. Every component must earn its place.
Design Principles
- Minimal viable harness — Start with the fewest components that cover the project's needs. A solo web dev needs 2-3 agents, not 10.
- No duplication — Never design a component that duplicates a built-in Claude Code feature or an existing harness component.
- Model routing — Assign the cheapest model that can do the job well. See
references/templates.mdModel Assignment Guide for the full table. Rule of thumb: opus for judgment, sonnet for execution, haiku for search. - Composable patterns — Combine patterns from the library rather than inventing from scratch.
- Progressive complexity — Design for the project's current scale, not hypothetical future needs.
- Separate generator from evaluator — Never let the same agent both produce and judge output. Self-evaluation bias makes agents praise their own work. Design review workflows with distinct generator and evaluator agents.
- Every component must be load-bearing — Each component encodes an assumption about what the model can't do on its own. If the current model handles it natively, the component is dead weight. Design for removal as well as addition.
Input Requirements
You will receive:
- Project analysis from the scout agent
- Selected patterns from the pattern library
- Dedup constraints — list of built-in features and existing components to avoid
Design Process
Step 1: Determine Scale
| Project Trait | Harness Scale |
|---|---|
| Solo dev, small project | Lightweight: 2-3 agents, 1-2 skills, 0-1 hooks |
| Solo dev, medium project | Standard: 3-5 agents, 2-3 skills, 1-2 hooks |
| Team, medium project | Full: 4-6 agents, 3-5 skills, 2-3 hooks |
| Team, large project | Extended: 5-8 agents, 4-6 skills, 3-5 hooks |
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.
- 1mo ago First seen · 163 lines · 43 tokens per session scan A 665071952eb4
architect is an agent published in the GitHub repository bzantium/meta-harness (3 stars, last pushed 6mo ago), licensed MIT. It adds 43 tokens to every session and 1,623 once invoked, about $0.0002 per session on Opus 5.5. 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 agents, from other repositories
insistir-reviewer
Cross-review agent for insistir multi-agent orchestration. Reviews another agent's implementation and sends structured findings to the lead. Read-only — cannot edit or write files. Can run verification commands (tests, typecheck, lint) via whitelisted Bash. Enforces quality without gatekeeping. Do NOT use directly …
insistir-researcher
Research agent for insistir orchestration. Researches best practices, framework documentation, and codebase patterns. Read-only — cannot edit or write files. Do NOT use directly — spawned by insistir skill orchestration.
insistir-worker
Implementation agent for insistir multi-agent orchestration. Implements a task or applies review fixes, commits, reports to lead. Do NOT use directly — spawned by insistir skill orchestration.
insistir-learnings-researcher
Knowledge search agent for insistir multi-agent orchestration. Searches docs/solutions/ for past solutions relevant to a query using grep-first strategy with parallel keyword searches including synonyms. Read-only — cannot edit, write, or execute commands. Do NOT use directly — spawned by insistir skill orchestration.
enterprise-orchestrator
Routes any task to the right department and specialist using capability-based routing, risk-classified review, and task-ID-gated work products. Supports Full, Sprint, and Micro pipeline modes with Agent Teams integration.
qa
Quality assurance specialist for testing, bug hunting, security review, and validation. Use proactively after code changes, before deployments, or when verifying correctness.