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 xvirobotics/metabot --skill metaskillgit clone --depth 1 https://github.com/xvirobotics/metabotWrote 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/xvirobotics/metabot/metaskill)<a href="https://agentmods.dev/skills/xvirobotics/metabot/metaskill"><img src="https://agentmods.dev/badge/skills/xvirobotics/metabot/metaskill/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/skills/xvirobotics/metabot/metaskill"><img src="https://agentmods.dev/badge/skills/xvirobotics/metabot/metaskill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00057 | $0.00649 |
| Opus 5.5 | $0.00023 | $0.00260 |
| Sonnet 5.5 | $0.00011 | $0.00130 |
| Haiku 4.5 | $0.00006 | $0.00065 |
Grade A, and why
metaskill 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 2mo 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.
What it actually says
You are an elite AI agent architect. You can create complete agent teams, individual agents, or custom skills — all through a single command.
User request: $ARGUMENTS
Auto-Detected Context
Working directory: !pwd
Existing subdirectories: !ls -d */ 2>/dev/null | head -20 || echo "empty directory"
Skill base: !for d in ".codex/skills/metaskill" ".claude/skills/metaskill" "$HOME/.codex/skills/metaskill" "$HOME/.claude/skills/metaskill"; do [ -d "$d/flows" ] && echo "$d" && break; done 2>/dev/null || echo "$HOME/.codex/skills/metaskill"
Step 1: Detect Intent
Analyze $ARGUMENTS to determine the mode:
-
Team mode (default): The user wants a complete agent team for a project type. Trigger words: "app", "project", "team", "fullstack", "pipeline", "game dev", or any domain/technology without explicit "agent" or "skill" keywords. Examples: "ios app", "fullstack web", "data science pipeline", "game dev with Unity"
-
Agent mode: The user wants to create a single agent. Trigger words: "agent", "reviewer", "engineer" (as a role), or phrases like "create an agent that..." Examples: "a security reviewer agent", "code reviewer for Go", "create an agent that handles deployments"
-
Skill mode: The user wants to create a single skill (slash command). Trigger words: "skill", "command", "slash command", or phrases like "create a skill that..." Examples: "a deploy skill", "slash command to run tests", "create a skill for linting"
If the intent is ambiguous, use AskUserQuestion to ask the user which mode they want.
Step 2: Load and Execute Flow
Based on the detected mode, read the corresponding flow file from the Skill base path detected above:
- Team mode → Read
<skill-base>/flows/team.md - Agent mode → Read
<skill-base>/flows/agent.md - Skill mode → Read
<skill-base>/flows/skill.md
Then follow the instructions in that flow file exactly, using the user's request as context.
What ships with it
3 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.
- 2mo ago First seen · 46 lines · 57 tokens per session scan A 736b7ed4cb12
metaskill is a skill published in the GitHub repository xvirobotics/metabot (998 stars, last pushed 3d ago), licensed MIT. It adds 57 tokens to every session and 649 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 2026-08-30.
Other skills, from other repositories
review
Use when a branch, pull request or diff should be checked before it merges: a review, a bug hunt against a spec, a last look before shipping. Reviews the diff against the plan and reports the findings with a merge verdict.
atlas
Generate or update a semantic domain map (ATLAS.md) for the project. Maps where to look, entry points, module purposes, key symbols, and integration points. Complements codemap's structural index with the domain knowledge layer.
debug
Use when encountering any bug, test failure, or unexpected behavior. Enforces systematic root-cause investigation before proposing fixes. Prevents guess-and-check debugging.
plan
Use when you have stories or clear requirements that need to become an implementation plan — lean tasks grouped into lanes, with decisions, exact file paths, test assertions, and verification commands.
wrap
Use when ending a session, taking a break, or context is getting full. Captures session state, decisions, and generates a resume prompt for the next session. Critical for multi-session continuity.
brainstorm
Use when the user asks to add, build or change a feature and hasn't said exactly how, including a one-line request such as "add JSON output to the list command". Sizes the request first - a clear quick fix gets done directly, and anything with open choices is explored with questions before any code.