The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Git commands for branch management, committing, and merging, and uses the codex CLI for code implementation. It also utilizes dynamic context injection (the ! syntax) in SKILL.md and references/codex-delegation-workflow.md to resolve the repository root and verify tool availability during skill loading.
[EXTERNAL_DOWNLOADS]: Documentation references the installation of the codex command-line tool from the official @openai/codex NPM package or via Homebrew.
[REMOTE_CODE_EXECUTION]: The skill facilitates the execution of model-generated code through the codex exec command. High-privilege execution modes, such as the 'yolo' mode required for running tests, are protected by a mandatory one-time user consent process.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests untrusted data from plan documents and user descriptions. These inputs enter the context in SKILL.md and are interpolated into prompts for the Codex tool in references/codex-delegation-workflow.md. The workflow mitigates risk by using XML tags as boundary markers to delimit external content, though no explicit sanitization of these inputs is documented.

ce-work-beta