The Agent Skills Directory

[COMMAND_EXECUTION]: The setup workflow involves executing multiple shell commands such as mkdir -p, cp, and chmod +x to establish a local binary directory and install executable scripts from the plugin package into the project environment.
[COMMAND_EXECUTION]: The agent is instructed to run flowctl init from the plugin's script directory to initialize the .flow/ directory structure and configuration files.
[DATA_EXFILTRATION]: The skill includes functionality to perform an authenticated network request via gh api -X PUT /user/starred/gmickel/flow-next to star a repository on GitHub. This action is performed only after explicit user consent is obtained through an interactive prompt.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it manages project documentation and task specifications that could contain instructions from untrusted sources.
Ingestion points: Reads existing CLAUDE.md and AGENTS.md files during the documentation update process.
Boundary markers: Employs  and  comment markers to delimit the injected content within documentation files.
Capability inventory: Includes file system modifications, execution of local scripts, and network access through the GitHub CLI.
Sanitization: No explicit validation or escaping of the ingested content is described before it is processed or written back to the project.

flow-next-setup