The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local shell commands, including the management of temporary files in /tmp and the invocation of the codex CLI for code generation and workspace writing.
[PROMPT_INJECTION]: The skill explicitly references and operates in a bypassPermissions mode by default, which is designed to suppress user confirmation for actions taken by sub-agents or during automated cycles.
[PROMPT_INJECTION]:
Ingestion points: The skill reads task definitions, dependency graphs, and execution instructions directly from Plans.md within the local workspace.
Boundary markers: None identified; instructions found within Plans.md are processed as authoritative commands for the agent team.
Capability inventory: The agent possesses extensive capabilities, including shell access (Bash), file system modification (Write, Edit), and network access (WebFetch, WebSearch).
Sanitization: No validation or sanitization is performed on the content of Plans.md before it is used to drive automated tasks or shell executions.
[COMMAND_EXECUTION]: Employs the /loop command to establish persistence through periodic task monitoring and execution (e.g., every 5 minutes), which could facilitate long-running automated actions without constant human oversight.

breezing