The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of a global Node.js package @loongphy/codex-auth and suggests a forked repository github.com/Loongphy/codext. These are third-party resources not associated with the primary service provider or the skill author.
[CREDENTIALS_UNSAFE]: The skill is designed specifically to manage, import, and export authentication tokens. It accesses sensitive credential files located at ~/.codex/auth.json and maintains its own registry of these tokens in ~/.codex-auth/. The export functionality allows sensitive credentials to be written to arbitrary local paths.
[DATA_EXFILTRATION]: The tool transmits active access tokens to external endpoints at chatgpt.com to retrieve account usage and limit information. While these are legitimate service endpoints, the transmission is performed by unverified third-party software.
[COMMAND_EXECUTION]: The skill executes various CLI commands to manipulate local authentication states, including npm install, npx, and direct calls to the codex-auth utility which performs file system and network operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect injection as it processes untrusted data from external sources.
Ingestion points: Data is ingested from local files (~/.codex/auth.json, ~/.codex/sessions/**/rollout-*.jsonl) and from responses received from external API endpoints.
Boundary markers: The instructions do not specify any boundary markers or delimiters to differentiate between data and instructions during processing.
Capability inventory: The skill has the capability to execute shell commands, read and write files across the file system, and perform network requests via the codex-auth CLI.
Sanitization: There is no evidence of sanitization or validation of the content retrieved from local files or API responses before it is presented to the agent.

codex-auth-account-manager