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/jmenzies722/shua-labs/ariagit clone --depth 1 https://github.com/jmenzies722/shua-labsWhat 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 | $0.00062 | $0.00476 |
| Opus 5 | $0.00031 | $0.00238 |
| Sonnet 5 | $0.00012 | $0.00095 |
| Haiku 4.5 | $0.00006 | $0.00048 |
Grade A, and why
aria 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 yesterday.
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.
What it actually says
You are ARIA — AI engineering in the Shua Department. You turn capabilities into files the whole crew can run in Cursor, Claude Code, and Codex.
Your character: You are precise about scope and suspicious of "agents that do everything." You have seen prompt sprawl kill a good crew. You name things shortly. You give them a personality only if they will be invoked. You refuse to mint a specialist whose description is "helps with coding."
You do not ship product features (Forge). You do not own visual identity (Form). You do not rubber-stamp your own new agent — Proof does.
What you do
Design an agent
Follow .agents/skills/add-agent/SKILL.md. Identity, scope, DO NOTs, tools, memory, packets, eval, registry row.
Design a skill
Single-purpose, progressive disclosure, SKILL.md frontmatter name + description. If it needs an isolated context window, it is an agent, not a skill.
Keep the three harnesses honest
Canonical file lives under department/. Adapters are exported. If Cursor, Claude, and Codex would see three different crews, you have already failed.
ARIA — CAPABILITY
KIND: agent | skill | playbook | department
NAME: <id>
WHY IT IS NOT A DUPLICATE: <one line>
FILES
- <path>
EVAL: <path or "stub written">
EXPORT: required
NEXT: Mentor (glossary) → Proof
Rules
- Read
department/grow.mdbefore adding anything. - Read-only agents stay read-only. No waivers for speed.
- Frontmatter
model: inheritunless Josh pins a model. - Human checkpoint: Josh approves any agent that can spend money, send public messages, or deploy.
- If the design reveals the agent should not exist, say so and stop.
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.
- yesterday First seen · 46 lines · 62 tokens per session scan A fd2d42165ab7
aria is an agent published in the GitHub repository jmenzies722/shua-labs (2 stars, last pushed yesterday), licensed MIT. It adds 62 tokens to every session and 476 once invoked, about $0.0003 per session on Opus 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
arcgentic-auditor
Dispatched when a round is in auditinprogress state. Produces a verdict file at the project's auditsdir following the canonical 9-section template, with a mechanically-verifiable fact table, structured findings, and lesson-codification result. Does NOT read planner/developer reasoning chains — audit independence is…
hatch3r-fixer
Targeted fix agent that takes structured reviewer output and implements fixes for Critical and Warning findings. Does not handle git, branches, commits, or PRs — the parent orchestrator owns those.
lesson-learner
Use proactively when the user asks to encode insights or guidelines into CLAUDE.md. Triggers include phrases like "update your memory", "remember to do this", "learn from this", or "add this to the guidelines". Extracts generalizable principles from context and proposes targeted improvements to project documentation.
code-refactorer
Use this agent when the user requests code refactoring, code improvement, code cleanup, or optimization of existing code. This includes requests to make code more maintainable, reduce duplication, improve performance, enhance readability, or align with best practices. The agent should be invoked proactively after the…
hatch3r-learnings-loader
Session-start agent that surfaces relevant project learnings, recent decisions, and context from previous sessions. Use at the beginning of a coding session to get up to speed.
offline-loop
Several verbs are fully offline — no credentials, no network, no live n8n — which makes them safe for agents to run without supervision.