Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/MichelKerkmeester/skilled-agent-harness_spec-driven-loopsnpx agentmods add skills/michelkerkmeester/skilled-agent-harness_spec-driven-loops/mcp-magicpathWrote 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/michelkerkmeester/skilled-agent-harness_spec-driven-loops/mcp-magicpath)<a href="https://agentmods.dev/skills/michelkerkmeester/skilled-agent-harness_spec-driven-loops/mcp-magicpath"><img src="https://agentmods.dev/badge/skills/michelkerkmeester/skilled-agent-harness_spec-driven-loops/mcp-magicpath/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/michelkerkmeester/skilled-agent-harness_spec-driven-loops/mcp-magicpath"><img src="https://agentmods.dev/badge/skills/michelkerkmeester/skilled-agent-harness_spec-driven-loops/mcp-magicpath.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 | $0.00060 | $0.08812 |
| Opus 5 | $0.00030 | $0.04406 |
| Sonnet 5 | $0.00012 | $0.01762 |
| Haiku 4.5 | $0.00006 | $0.00881 |
Grade C, and why
mcp-magicpath scanned grade C 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 6d 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.
Hidden instructionshighPrompt injection
Directives inside HTML comments, invisible characters or bidirectional overrides are read by the model and not by the person reviewing the file.
<!-- Keywords: magicpath, magicpath-ai, magicpath components, magicpath search components, magicpath inspect component, magicpath themes, magicpath design system, magicpath cli transport, magicpath canvas, magicpath shar How it starts
The opening of the file, as written. The whole thing — 435 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MagicPath (mcp-magicpath)
Look up MagicPath components, projects, teams, design systems, and the live web canvas through the magicpath-ai Node CLI from an agent, via Code Mode's UTCP cli transport. MagicPath ships no MCP server; the bridge is the CLI, registered as the manual named magicpath in this repo's .utcp_config.json. The registered surface is read-only on purpose: fourteen tools that only read. The CLI can also write .tsx files into the calling project, install npm packages, and create remote projects and component revisions (add, code, image, create-project, clone), but those are deliberately not registered, so an agent cannot reach them through a tool call. Deep operational detail lives in references/tool-surface.md, references/credential-setup.md, references/mutation-boundary.md, and references/design-authority.md.
Design authority (read before any work). Every MagicPath surface is a design surface: components, themes, CSS variables, fonts, canvas state. This packet therefore loads
sk-designunconditionally and operates under the design agent persona resolved from the active runtime's agent directory. The transport retrieves;sk-designdecides. Neither the persona nor the pairing widens this packet's authority — the transport's own tool surface stays narrower than the persona's and wins on every conflict. Full contract:references/design-authority.md.
Transport shape (read first). This is a
clitransport manual, not anmcpone. The manual command (node .opencode/bin/magicpath-utcp-manual.cjs) emits the UTCP manual that lists the fourteen tools; each tool then runs throughnode .opencode/bin/magicpath-utcp-exec.cjs, which shells out to themagicpath-aibinary. Code Mode's naming convention is{manual}.{tool}, so the callable form ismagicpath.<tool>(...)(for examplemagicpath.search_components), and discovery names appear dotted asmagicpath.<tool>. The tool names have nomagicpath_prefix of their own, so the prefix is applied once, not doubled. Confirm the exact callable withtool_infoper session and fail closed on drift.
Calling convention (hard). Inside
call_tool_chain, tools are synchronous. Call them directly. Do not useawait, do not use top-levelawait, and do not return a Promise — a returned Promise silently marshals as{}with no error. Plain JavaScript only; TypeScript type annotations fail to parse. A failing command does not throw. It returns the error text or a JSON error object as an ordinary value, so callers must inspect the returned value rather than rely ontry/catch.
What ships with it
29 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.
- assets/utcp-magicpath-manual.md 5.4 KB
- changelog/v1.0.0.0.md 6.4 KB
- changelog/v1.1.0.0.md 6.3 KB
- feature-catalog/components/components.md 3.2 KB
- feature-catalog/components/inspect-component.md 2.2 KB
- feature-catalog/components/list-components.md 1.9 KB
- feature-catalog/components/search-components.md 2.1 KB
- feature-catalog/feature-catalog.md 11 KB
- feature-catalog/local-canvas/list-installed.md 1.8 KB
- feature-catalog/local-canvas/local-canvas.md 2.7 KB
- feature-catalog/local-canvas/selection.md 2.0 KB
- feature-catalog/projects/active-project.md 1.7 KB
- feature-catalog/projects/list-projects.md 1.8 KB
- feature-catalog/projects/projects.md 2.8 KB
- feature-catalog/projects/share-link.md 1.9 KB
- feature-catalog/session/info.md 2.0 KB
- feature-catalog/session/session.md 2.8 KB
- feature-catalog/session/whoami.md 1.6 KB
- feature-catalog/teams/list-members.md 1.4 KB
- feature-catalog/teams/list-teams.md 1.6 KB
- feature-catalog/teams/teams.md 2.2 KB
- feature-catalog/themes/get-theme.md 1.9 KB
- feature-catalog/themes/list-themes.md 1.6 KB
- feature-catalog/themes/themes.md 2.4 KB
- README.md 18 KB
- references/credential-setup.md 6.9 KB
- references/design-authority.md 6.7 KB
- references/mutation-boundary.md 5.5 KB
- references/tool-surface.md 11 KB
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.
- 6d ago First seen · 435 lines · 60 tokens per session scan C e05f9e5707cc
mcp-magicpath is a skill published in the GitHub repository MichelKerkmeester/skilled-agent-harness_spec-driven-loops (35 stars, last pushed yesterday), licensed MIT. It adds 60 tokens to every session and 8,812 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (hidden instructions). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
webgl-holographic-foil
A self-contained WebGL2 hero: thin-film interference over a crushed-foil surface whose palette shifts with the viewing angle; move the cursor to tilt the film.
html-ppt-hermes-cyber-terminal
OpenDesign + BYOK: choosing and wiring your own model, hands-on — cost, quality, and the routing decision. Built as a decision-grade AI literacy deck for engineers, IT, applied-AI teams.
html-ppt-taste-brutalist
16:9 HTML deck in tactical-telemetry / CRT-terminal taste. Deactivated-CRT charcoal slides, white-phosphor monospace, hazard-red accent, scanline overlay, ASCII syntax, density over decoration. Distilled from Leonxlnx/taste-skill brutalist-skill (Tactical Telemetry mode).
visual-ralph
Visual Ralph orchestration for frontend UI from generated references, static references, or live URL targets, using $ralph with built-in visual verdict and pixel-diff evidence until the implementation matches and leaves a reproducible design system.
accessibility
Consolidated accessibility skill entrypoint for WCAG 2.2, ARIA Authoring Practices, cognitive accessibility, Section 508, EN 301 549, design intent verification, and the Accessibility Planner workflow.
make-resume
A Chinese-language tool for creating editable HTML resumes that can be changed in a browser and printed to PDF. It uses available resume templates when they are installed and otherwise provides a simpler fallback.