The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is built around executing shell commands via the codex-cli. It spawns background processes using codex exec and utilizes the macOS caffeinate utility to manage system power states during execution. It also performs environmental checks using which codex and codex --version.
[EXTERNAL_DOWNLOADS]: The skill requires the third-party codex-cli (specifically referencing version 0.138.0) to be installed and configured on the host machine. This external dependency is critical for the skill's operation but is not a standard system utility.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from multiple sources.
Ingestion points: The agent reads workspace configuration from AGENTS.md and processes output files generated by Codex workers in the out/ directory.
Boundary markers: Task instructions are structured using a specific 'Brief' template to separate job specifications from the wider context.
Capability inventory: The agent has the capability to execute shell commands (codex exec) and perform read/write operations within the local workspace.
Sanitization: The skill implements a mandatory review gate where the agent validates worker outputs against pre-defined acceptance criteria, reducing the risk of automatically following instructions embedded in worker output.

suede-codex-fleet