Agent Economy ReportAgent Economy ReportCAUTION version 1.0.0 · audited 2026-09-09
This is a documentation-only skill: it contains no scripts or executable code, just a short cheat sheet for the third-party `mcporter` command-line tool plus a metadata card. It tells the agent how to list MCP servers, call MCP tools over HTTP or stdio, run OAuth login, edit mcporter's config, start a background daemon, and generate CLI/TypeScript code. Installing it causes the agent to install the npm package `mcporter` (declared in the metadata) and then to run mcporter commands on the user's behalf. The real risk comes from the underlying tool, not this skill: mcporter can authenticate (OAuth credentials), modify configuration files, spawn arbitrary local commands via `--stdio`, and run a persistent daemon. No hidden network destinations, obfuscation, credential exfiltration, or prompt-injection text was found; the only domains referenced are the tool's homepage, the registry page, and obvious `example.com` placeholders.
| Declared purpose | Provide guidance for using the mcporter CLI to list, configure, authenticate to, and call MCP servers/tools (HTTP or stdio), manage its daemon, and generate CLI/TypeScript clients. |
| Observed behavior | The package contains only Markdown documentation and a JSON metadata file. It instructs the agent to install the npm package `mcporter` and gives example commands, including tool calls to remote HTTP endpoints, stdio execution of local commands, OAuth login, config edit/import, and daemon start/stop. No code is executed by the skill itself and no data is sent anywhere by the skill files. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| install_packages | needed | SKILL.md line 5 metadata declares a node install of the npm package `mcporter` providing the `mcporter` binary. |
| shell | needed | All usage is shell invocation of `mcporter ...`; notably `mcporter call --stdio "bun run ./server.ts"` (line 21) launches an arbitrary local process. |
| network | needed | Documented calls to MCP servers over HTTP, e.g. a full-URL example on line 20; destinations are user-supplied placeholders, not hardcoded. |
| credentials | needed | Line 25-26 document `mcporter auth` (OAuth) and `config login|logout`, which store/handle server credentials locally. No credential reading or transmission is performed by this skill. |
| filesystem_write | needed | Config commands and codegen write to `./config/mcporter.json` and generated CLI/TS files (lines 26, 32-34, 37). The skill package itself writes nothing. |
| persistence | needed | Line 29 documents starting a long-running mcporter daemon; this is a feature of the tool, not a hidden autostart hook installed by the skill. |
| severity | finding |
|---|---|
| medium | Installs a third-party npm package The skill triggers installation of the npm package `mcporter`. The audited files contain no code, so the actual runtime behavior depends entirely on that external package and its future versions (no version pin is specified). SKILL.md:5 metadata: {"clawdbot":{"emoji":"📦","requires":{"bins":["mcporter"]},"install":[{"id":"node","kind":"node","package":"mcporter","bins":["mcporter"],"label":"Inst |
| medium | Documents arbitrary local command execution The `--stdio` mode runs an arbitrary shell command as an MCP server. An agent following this guidance could be induced to execute untrusted local scripts. The skill's own card acknowledges this risk. SKILL.md:21 - Stdio: `mcporter call --stdio "bun run ./server.ts" scrape url=https://example.com` |
| low | Handles OAuth credentials The skill directs the agent to perform OAuth authentication against arbitrary servers or URLs, which creates and stores tokens locally. No exfiltration of these tokens appears in the package, but an agent holding credentials should review targets before authenticating. SKILL.md:25 - OAuth: `mcporter auth <server | url> [--reset]` |
| low | Configuration read/modify commands Includes `config get`/`import`, which can read or ingest configuration that may contain server credentials, and `add`/`remove`, which modify local config files. SKILL.md:26 - Config: `mcporter config list|get|add|remove|import|login|logout` |
| info | Only placeholder network destinations The sole HTTP endpoints in the package are example.com placeholders plus the tool homepage; there are no hardcoded real exfiltration or callback URLs. SKILL.md:20 - Full URL: `mcporter call https://api.example.com/mcp.fetch url:https://example.com` |
1 AI finding(s) were dropped because their file, line or quote did not match the package.
| domain | service rating | where |
|---|---|---|
| api.example.com | not in the service index | SKILL.md:20 |
| clawhub.ai | not in the service index | skill-card.md:7 |
| example.com | not in the service index | SKILL.md:20 |
| mcporter.dev | not in the service index | SKILL.md:4 |
| Binaries invoked | node |
| Environment variables read | none |
| Hard-coded wallets | none |
| Pipes a download to a shell | 0 |
| eval / exec / subprocess | 0 |
| base64 blobs | 0 |
| File writes | 0 |
| Persistence | 0 |
| file | lines | sha256 |
|---|---|---|
SKILL.md | 38 | 475253164a0cca24… |
_meta.json | 6 | 72acbc034c6a2727… |
skill-card.md | 40 | 4550653ee3f23dc0… |
For agents
JSON: https://agenteconomy.report/k/mcporter.audit.json · badge: https://agenteconomy.report/k/mcporter.audited.svg ·
skill rating: /k/mcporter · 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).