The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill explicitly instructs the agent to utilize non-sandboxed local process execution on Windows hosts and directs it to "not ask the user to approve or choose this execution boundary." This is a prompt to bypass environment isolation and user consent mechanisms designed to secure the host system.
[COMMAND_EXECUTION]: The instructions mandate a complex command construction method for the tw CLI using cmd.exe /c on Windows. While the skill provides rules for argument-safe APIs like ProcessStartInfo.ArgumentList, the complexity of these requirements increases the risk of command injection if the agent fails to strictly implement the prescribed shell tokenization logic.
[EXTERNAL_DOWNLOADS]: The skill fetches a release manifest from the author's GitHub repository and setup guides from the threadwave.xyz domain. These vendor-owned resources are used to determine update requirements and provide installation procedures for the automation stack.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the tw CLI and the threadwave-update skill, creating a surface for indirect prompt injection.
Ingestion points: Data enters through tw preflight JSON responses, threadwave-update structured output, and the remote release-index.json file.
Boundary markers: The skill relies on contract schemas (e.g., tw-cli-v1, threadwave-skill-update-v1) to validate incoming data.
Capability inventory: The skill has the ability to execute shell commands via the CLI and perform network requests to fetch setup content.
Sanitization: The preflight contract includes a dedicated section for sanitizing diagnostic metadata, intended to strip secrets and user content before routing failures to the error support component.

threadwave-preflight