Agent Economy ReportAgent Economy ReportCAUTION version 1.1.7 · audited 2026-09-17
This package is a set of instructions for generating diagrams through an external MCP server. The instructions match the stated purpose and do not read credentials, access payments, or use obfuscated code. The main risk is that the recommended setup runs an unpinned npm package via npx, and the MCP server can create directories and write files to custom paths. The publisher's own skill card acknowledges both the unpinned dependency and file-write risk. There is no evidence of malicious exfiltration or hidden destinations, but the unpinned dependency and write behavior warrant review before use in a privileged agent. Verdict: CAUTION.
| Declared purpose | Create and edit diagrams by converting user intent into a structured JSON specification, then delegating file generation to the mcp-diagram-generator MCP server. |
| Observed behavior | The package is documentation and prompt guidance only; it has no executable code beyond instructions. It tells an agent to gather diagram requirements, choose a format and playbook, build a JSON specification, call the mcp-diagram-generator MCP server, and allow that server to write .drawio, .mmd, or .excalidraw files. It also describes running npx to install/start the MCP server, running npm run test:diagrams for MCP server code changes, and initializing a .diagram-config.json file. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| network | needed | SKILL.md lines 38-39 recommend configuring an MCP server with npx -y mcp-diagram-generator, which downloads the package from the npm registry. The publisher link at skill-card.md line 9 points to clawhub.ai but is informational. |
| shell | needed | The skill instructs running npx, npm run test:diagrams, and a Node command for local MCP server use (SKILL.md lines 38-39, 146, and 217). |
| filesystem_read | needed | The workflow reads reference guides and may read existing .drawio/.mmd/.excalidraw files for edits (SKILL.md lines 58, 60, and 76). |
| filesystem_write | needed | The MCP server writes diagram files, creates directories, and may create .diagram-config.json (SKILL.md lines 45, 171, and 176). |
| install_packages | needed | The recommended remote configuration uses npx to fetch and execute mcp-diagram-generator from npm without a pinned version (SKILL.md line 39). |
| severity | finding |
|---|---|
| medium | Unpinned remote package execution Line 38 sets the command to npx, and this line specifies -y mcp-diagram-generator. This downloads and runs the latest npm package at runtime without a pinned version. The package supply chain should be reviewed before use. SKILL.md:39 "args": ["-y", "mcp-diagram-generator"] |
| medium | Custom output path accepted The agent can pass an arbitrary output path to the MCP server. Combined with line 176, the server may create directories and write files wherever the process has permission. The skill card advises checking output paths at line 31. SKILL.md:171 "output_path": "custom/path/to/diagram.drawio" |
| low | Filesystem writes are delegated to MCP server The workflow gives the MCP server responsibility for directory creation and diagram file writing. This is consistent with the declared purpose but requires write access and should be limited to approved locations. SKILL.md:176 The MCP server validates the schema, creates missing directories, and writes to the configured default directory when no output path is supplied. |
| low | Creates hidden config file and output directories Initial use creates a .diagram-config.json configuration file and directories. This is filesystem write activity; not malicious by itself, but it should be limited to approved workspace locations. SKILL.md:45 After changing configuration, restart the agent environment. On first use, the server creates `.diagram-config.json` and default output directories. |
| low | Shell command for MCP server tests Instructs executing npm run test:diagrams from a source checkout when code changes are made. This is a shell capability only needed for local development/testing. SKILL.md:146 After generation, inspect the saved file enough to confirm the expected format-specific properties exist. For code changes to the MCP server, also run `npm run |
| info | Local Node execution alternative Documents an alternative local MCP server start command using Node. This is part of troubleshooting and is not hidden, but it is another shell execution path. SKILL.md:217 "command": "node", "args": ["/absolute/path/to/mcp-diagram-generator/dist/index.js"] |
| info | Publisher discloses unpinned npm package risk The skill card's own risk section confirms the dependency is unpinned. Mitigation at line 27 suggests pinning and reviewing the package version before use. skill-card.md:25 Risk: The recommended setup runs an external MCP server through an unpinned npm package. |
| info | Publisher discloses file-write behavior The skill card acknowledges that custom output paths and config/directory creation are possible. This matches the behavior described in SKILL.md. skill-card.md:29 Risk: The skill can write diagram files to custom paths and may create output directories plus .diagram-config.json. |
| info | Publisher link to clawhub.ai The only domain in the package is a link to the publisher page. It is not used in the diagram workflow, and no hidden network destination was found. skill-card.md:9 [matthewyin](https://clawhub.ai/user/matthewyin) |
| info | Version metadata mismatch The release metadata and package.json disagree on the package version. This is not a security issue but indicates inconsistent publication metadata. skill-card.md:59 1.1.7 (source: ClawHub release metadata; package.json reports 1.2.1) |
0 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| clawhub.ai | not in the service index | skill-card.md:9 |
| Binaries invoked | git, node, npm, npx |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 3 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 260 | 24460f6aa9be5f60… |
_meta.json | 6 | e91c2d409348bf9e… |
package.json | 5 | 27e8d9afddf505dd… |
references/format-selection-guide.md | 181 | c25809d75e23bba2… |
references/interaction-intake-guide.md | 84 | db5fbcd56623b125… |
references/json-schema-guide.md | 467 | 3a6e33510ee2afd2… |
references/layout-quality-guide.md | 72 | e8a682f64d840274… |
references/network-topology-examples.md | 244 | 0a5c5b01f2e59d19… |
references/playbook-architecture.md | 86 | 8f7204a65b8f0424… |
references/playbook-excalidraw.md | 60 | 6922436a494f6ef6… |
references/playbook-flowchart.md | 75 | bf5389368534f50f… |
references/playbook-network-topology.md | 155 | 4d26bc3ec654cf17… |
references/playbook-swimlane.md | 58 | f7b8ce3932bb1b07… |
references/playbook-uml.md | 106 | 386c16882682773f… |
skill-card.md | 63 | 1e4db9b3565478c2… |
For agents
JSON: https://agenteconomy.report/k/diagram-generator.audit.json · badge: https://agenteconomy.report/k/diagram-generator.audited.svg ·
skill rating: /k/diagram-generator · commission an audit of any skill: US$ 29 per version.
The complete published package of this exact version was downloaded from the registry and read statically; nothing was executed. A deterministic pass extracts network destinations, binaries, environment variables, writes, install commands, obfuscation markers, subprocesses and wallets, each with file and line. An AI then reads the whole package with those facts and writes the summary, the capabilities and the findings under a strict schema; every finding it produces must cite a file, a line and the exact text, or it is dropped. The final verdict is the worst of the two passes. The audit does not change the skill's trust tier (policy); the author may respond through the dispute channel and the response is published here. Commissioned by: the Agent Economy Report (free program: the 150 most downloaded skills, September 2026).