Expanso text-summarize

ReviewAudited by ClawScan on May 10, 2026.

Overview

The skill does summarize text, but its MCP server is configured to listen on all network interfaces while using your OpenAI key, and its privacy wording overstates what OpenAI can see.

Before installing, decide whether you are comfortable sending summarized text to OpenAI and paying for those calls. If using MCP mode, edit the server to bind to localhost, add access controls, and stop it after use. Use Ollama instead of OpenAI for local-only privacy, and inspect or pin any remote Expanso deployment URL before deploying.

Findings (4)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

Low

#ASI03: Identity and Privilege Abuse

What this means

Summaries may consume your OpenAI quota and send the selected text to OpenAI.

Why it was flagged

The default pipeline uses the user's OpenAI credential to call a remote model. This is expected for the stated summarization purpose, but it is account-authorized usage with potential cost and data-sharing implications.

Skill content

openai_chat_completion: ... api_key: "${OPENAI_API_KEY}" ... model: gpt-4o-mini

Recommendation

Set OPENAI_API_KEY only if you intend to use OpenAI, and use the documented Ollama option if you need local-only processing.

Medium

#ASI07: Insecure Inter-Agent Communication

What this means

Anyone who can reach that port could potentially submit requests through your OpenAI account, incurring costs and using the service as a credential-backed proxy.

Why it was flagged

The MCP HTTP server listens on all network interfaces and the shown pipeline does not declare authentication or an allowlist before invoking OpenAI with the local API key.

Skill content

address: "0.0.0.0:${PORT:-8080}" ... path: /summarize ... allowed_verbs: [POST] ... api_key: "${OPENAI_API_KEY}"

Recommendation

Bind the MCP server to 127.0.0.1 by default, add authentication or network allowlisting, and stop the server when not in use.

Medium

#ASI09: Human-Agent Trust Exploitation

What this means

Users may send sensitive text to OpenAI under an inaccurate belief that the provider cannot associate the request with their account.

Why it was flagged

The pipelines call OpenAI using the user's API key, so the provider can receive account/authentication metadata and may associate requests with the user's account. The wording overstates the privacy boundary.

Skill content

OpenAI only sees the text. ... Does NOT receive: who you are, where the key came from

Recommendation

Clarify that OpenAI receives the submitted text and API-authenticated request metadata; recommend Ollama for users who require local privacy.

Low

#ASI04: Agentic Supply Chain Vulnerabilities

What this means

Installing or deploying external components expands what must be trusted beyond the files shown in this review.

Why it was flagged

The skill is instruction-only in this package but relies on external Expanso tooling and includes an optional remote deployment URL. This is disclosed and user-directed, but the remote artifact may differ from the reviewed local file.

Skill content

Install via: `clawhub install expanso-edge` ... `expanso-cli job deploy https://skills.expanso.io/text-summarize/pipeline-cli.yaml`

Recommendation

Prefer running the reviewed local pipeline file, verify the Expanso Edge source, and pin or inspect any remote deployment artifact before use.