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 skills add bzantium/meta-harness --skill create-harnessgit 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/skills/bzantium/meta-harness/create-harness)<a href="https://agentmods.dev/skills/bzantium/meta-harness/create-harness"><img src="https://agentmods.dev/badge/skills/bzantium/meta-harness/create-harness.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 | $0.00117 | $0.03735 |
| Opus 5 | $0.00059 | $0.01868 |
| Sonnet 5 | $0.00023 | $0.00747 |
| Haiku 4.5 | $0.00012 | $0.00374 |
Grade B, and why
create-harness scanned grade B 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 7d 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
Add hooks to `{project}/.claude/settings.json`: How it starts
The opening of the file, as written. The whole thing — 388 lines — stays where its author put it; the contents beside it link to each section on GitHub.
meta-harness — Intelligent Harness Creator
Generates project-specific Claude Code harnesses by analyzing the codebase, selecting proven patterns from a curated library, and producing agents, skills, hooks, and rules — while auto-checking official features to avoid duplication.
Workflow
Phase 0: Deduplication Pre-check
Before any design work, establish what already exists:
-
Read the built-in features reference — Load
references/official-builtins.mdto know what Claude Code provides natively (50+ commands, 25 hook events, 5 subagent types, Agent Teams, etc.) -
Live update check — Use WebSearch to query "Claude Code changelog latest features" to catch anything newer than the reference. If new built-in features are found that overlap with what you'd generate, note them.
-
Check target project — Look for existing
.claude/directory contents:.claude/agents/— existing agent definitions.claude/skills/— existing skill definitions.claude/settings.json— enabled plugins, hook configsCLAUDE.md— existing project rules.mcp.json— MCP server configs
-
Compile exclusion list — Everything found above is off-limits for generation. Present the list to the user.
If the user requests something that's already built-in, inform them and suggest the native feature instead.
Phase 0.5: Clarification (if needed)
Before analysis, evaluate whether the user's request provides enough context. Use AskUserQuestion to gather missing information if:
- The project is empty/new and the user didn't describe what they're building
- The user's description is vague (e.g., "harness this project" with no context)
- Key decisions affect harness design (team size, deployment target, timeline)
Example questions to ask (use the system popup, NOT text):
- "What is the main purpose of this project?" (with options based on what you can infer)
- "What's the team size?" (Solo / Small 2-5 / Medium 6-15 / Large 15+)
- "What's the primary workflow?" (Feature development / Bug fixes / Research / Content creation)
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 7d ago First seen · 388 lines · 117 tokens per session scan B 6d65b7afb958
create-harness is a skill published in the GitHub repository bzantium/meta-harness (2 stars, last pushed 5mo ago), licensed MIT. It adds 117 tokens to every session and 3,735 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
goal-loop
Loop engineering as gradient descent — a guided goal loop where an agent factory generates goal-specialized agents (implementer, verifier, diagnoser, judge) and iterates forward → loss → backward → update until the goal is provably met. Evidence is split into visible validation (the implementer's loss) and held-out…
metaprompt
Generate a complete, ready-to-use prompt for a target model and harness. Triggers: metaprompt, generate a prompt for, write me a prompt, create a system prompt, prompt engineer this, optimize this prompt.
codex-judge
Cross-provider LLM judge using Codex (GPT-5) to evaluate implementations written by Claude. Internal skill invoked by the insistir lead during the review loop. Not user-facing.
file-todos
File-based TODO lifecycle for tracking review findings, managing technical debt, and organizing work items. Each TODO is a standalone markdown file with YAML frontmatter, named by convention: {id}-{status}-{priority}-{description}.md. Status transitions drive the lifecycle: pending (created from review) -> ready…
agent-sdk-wizard
Build a working Claude Agent SDK agent by asking the user 9 illustrated questions, one at a time, then generating and running the code. Trigger on "crea un agente", "quiero hacer un agente con el SDK", "ayúdame a armar mi agente", "nuevo agente paso a paso", "agent sdk wizard", "build an agent with the Claude Agent…
image-to-frontend
This skill should be used when the user asks to "image to frontend", "build this UI", "make me a landing page", "design and ship this site", "generate a mockup", "turn this screenshot into code", or hands over a product brief or reference image and wants a working React or HTML page. Also covers single-shot app…