The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use standard Git commands (git log, git worktree) to manage the lifecycle of user-mode skill files and versioning. These are standard developer operations for local file management.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from agent-transcripts/ to identify user patterns.
Ingestion points: Reads existing agent chat histories from the workspace-scoped agent-transcripts/ directory.
Boundary markers: Explicitly restricts scanning to the current workspace path to prevent cross-project data access, although it lacks specific delimiters for the mined content.
Capability inventory: Performs file writing via the built-in create-skill tool and executes version control operations.
Sanitization: No specific sanitization or filtering of the transcript content is performed during the pattern extraction phase.
[SAFE]: The operations performed by the skill are local, transparently described, and include explicit instructions to prevent data leakage across different project workspaces. The external skill references (poteto-mode, unslop, create-skill) are treated as trusted platform-specific resources.

automate-me