The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill synchronizes source code from the developer's repository (code-yeongyu/lazycodex) and OpenAI's official repository (openai/codex) to /tmp. These downloads are used exclusively for auditing the local installation against authoritative sources.
[COMMAND_EXECUTION]: Executes a series of diagnostic commands, including git, gh (GitHub CLI), find, and version checks for the codex binary. These operations are limited to information gathering and are necessary for the skill's stated purpose.
[PROMPT_INJECTION]:
Ingestion points: Processes external data from GitHub issue lists and local command outputs to check for known bugs and configuration drift (SKILL.md).
Boundary markers: The skill does not explicitly use delimiters for external data, but the context is restricted to a structured report output.
Capability inventory: The environment allows file system reads, network access via official GitHub tools, and execution of the local codex binary.
Sanitization: Content retrieved from GitHub is included in the final report without specific sanitization; however, the risk is minimal given the technical and diagnostic nature of the output.
[DATA_EXPOSURE]: Accesses the ~/.codex/config.toml file to verify configuration integrity. While this file may contain application-specific settings, the access is scoped to the software being diagnosed.

lcx-doctor