PlanWright

The materials indicate no required keys or environment variables, and there is no request for API tokens, account credentials, or local sensitive authentication material; credential exposure appears limited.

It is known to connect to the remote endpoint mcp.planwright.tools. As a control-plane style MCP tool, task content, agent state, or audit-trail data may be transmitted through that endpoint; the materials do not describe data minimization, encryption details, or retention policy.

The system checks explicitly mark executes-code, and the description references autonomous software labor, indicating routine agent capability to launch local code or processes. This is a common MCP capability rather than a standalone red flag, but it should be constrained to a controlled workspace and low-privilege runtime.

The materials do not explicitly list which files, directories, or resources it can read or write; combined with its code-execution capability and control-plane role, it may in practice touch workspace contents, command output, and task context. There is no clear evidence of excessive permissions beyond its stated function, but the access boundary is not transparent.

Positive signals include an official registry source, an auditable open-source repository, and updates within the last year; these materially reduce supply-chain risk. Points to watch are the missing README, undeclared license, and 0 stars, which weaken public governance and community scrutiny, so source and dependency locking should still be reviewed.