The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates task completion by executing shell commands and delegating workflows to external command-line tools such as codex and harness-work.
[PROMPT_INJECTION]: The skill identifies task scopes by reading the contents of Plans.md, which creates a surface for indirect prompt injection if that file contains instructions from untrusted sources.
Ingestion points: Task definitions are ingested from the Plans.md file using the Read tool.
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands provided in this coordination layer.
Capability inventory: The skill operates with significant privileges, including the ability to use Bash, Write, and Edit tools.
Sanitization: The skill does not validate or sanitize task descriptions before passing them to the sub-agents for implementation.

breezing