io.github.CSOAI-ORG/accessibility-ai-mcp

The materials explicitly state that no keys or environment variables are required, and no API keys, tokens, or account credentials are mentioned; based on the available facts, credential leakage or misuse exposure appears low.

The materials explicitly state that there are no remote endpoint hosts, and they do not declare any external service connections or third-party data transmission; based on current information, no network-egress red flags are visible, though runtime monitoring is still advisable due to sparse documentation.

The system checks explicitly indicate executes-code capability, meaning it can run code/processes locally; this is a standard sensitive capability for MCP tools and warrants caution by default. Because the README is absent, the exact boundary of system capabilities it may invoke cannot be confirmed from the materials.

The materials do not specify which local files, directories, or resources it can read or write, so the data-access scope is not transparent; given its local execution capability, it should be treated with least-privilege controls. There is no explicit evidence of requesting permissions far beyond its stated purpose, so caution is more appropriate than risk.

Positive factors include an official registry source, a public open-source repository, and updates within the past year; these materially reduce supply-chain risk. Points of caution are the essentially missing README, undeclared license, and low community adoption (0 stars), which limit auditability and ecosystem validation.