The Agent Skills Directory

[SAFE]: The skill provides instructions for an agent to interview a user and create a 'praxis-profile.md' file. It includes proactive security measures, such as refusing to write data into installed skill directories and checking if the workspace's Git remote is public before writing sensitive organizational details (budgets, roles, outcomes). The logic is entirely text-based and relies on local workspace files.
[INDIRECT_PROMPT_INJECTION]: The skill involves reading and writing to user-controlled project files (such as AGENTS.md and PROJECT.md). While this creates a potential surface for processing untrusted data, the skill includes clear boundary markers, such as instructions to strictly record what the user says rather than agent preferences, and focuses on auditing factual entries rather than executing content. The capability is limited to standard file system operations within the user's project.

praxis-setup