Agent Economy ReportAgent Economy ReportCAUTION version 1.0.0 · audited 2026-09-17
This package is a short instruction card for the qmd command-line tool, which indexes and searches local files using BM25, vector, hybrid search, and MCP mode. It contains no executable code in the package itself beyond metadata that tells the agent to install qmd from GitHub. The declared commands read local directories, write an index under ~/.cache/qmd, and may contact an Ollama endpoint that defaults to localhost. No credential exfiltration, persistence, payment redirection, or hidden network destinations were found. The main caution is that the install source is an unpinned GitHub repository, which the skill card itself acknowledges as a supply-chain risk.
| Declared purpose | Local search/indexing CLI (BM25 + vectors + rerank) with MCP mode. |
| Observed behavior | The package provides CLI usage instructions. It tells the agent to install qmd via Node from https://github.com/tobi/qmd, then use commands to add local collections, update an index, query with BM25/vector/hybrid search, retrieve document snippets, and optionally start MCP mode. The index is stored under ~/.cache/qmd by default, and embeddings/rerank can use an Ollama URL that defaults to http://localhost:11434. |
| Verdict, rules only | SAFE |
| Verdict, AI | CAUTION |
| Final verdict | CAUTION (never better than either pass) |
| capability | vs purpose | detail |
|---|---|---|
| network | needed | Install metadata downloads qmd from https://github.com/tobi/qmd. Search may use Ollama at OLLAMA_URL, defaulting to http://localhost:11434. |
| shell | needed | The skill is used by running qmd CLI commands such as qmd collection add, qmd update, qmd search, qmd vsearch, qmd query, qmd get, and qmd mcp. |
| filesystem_read | needed | qmd indexes local files and retrieves document content from specified local paths. |
| filesystem_write | needed | Running indexing commands causes qmd to create and update a local index under ~/.cache/qmd by default. |
| install_packages | needed | Install metadata uses Node to install the qmd package from GitHub, supplying the qmd binary. |
| other | needed | MCP mode is declared: `qmd mcp` can run qmd as an MCP server for integration with AI tools. |
| severity | finding |
|---|---|
| medium | Unpinned GitHub package source The install metadata fetches the qmd package from an unpinned GitHub repository. The fetched code is not tied to a release or commit, which creates supply-chain risk. SKILL.md:5 https://github.com/tobi/qmd |
| medium | Supply-chain risk acknowledged The skill card itself flags the unpinned GitHub package source as a supply-chain risk and recommends installing from a reviewed release or pinned commit when available. skill-card.md:24 Risk: The install metadata references an unpinned GitHub package, creating supply-chain risk. |
| info | Declared local file indexing and search The skill directs the agent to index local files and run searches against that index. SKILL.md:10 Use `qmd` to index local files and search them. |
| low | Local index cache location Indexed content is cached under the current user's home directory and could become searchable by the agent or other local processes. SKILL.md:25 Index lives under `~/.cache/qmd` by default. |
| low | Localhost Ollama endpoint Vector/hybrid search and reranking may send data to a configurable Ollama endpoint, defaulting to localhost. SKILL.md:24 Embeddings/rerank use Ollama at `OLLAMA_URL` (default `http://localhost:11434`). |
| info | MCP server mode available qmd can run in MCP mode, exposing search/indexing operations as an MCP service. SKILL.md:26 MCP mode: `qmd mcp`. |
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 |
| github.com | not in the service index | SKILL.md:5 |
| tobi.lutke.com | 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 | 26 | 8d0a688885a61062… |
_meta.json | 6 | 22bea938c8aef5f9… |
skill-card.md | 54 | bb45f43d27e8c664… |
For agents
JSON: https://agenteconomy.report/k/qmd.audit.json · badge: https://agenteconomy.report/k/qmd.audited.svg ·
skill rating: /k/qmd · 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).