The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a local Python script (codex_bridge.py) located within the plugin's root directory to handle communication with an external reviewer model.
[DATA_EXFILTRATION]: Project artifacts, including manuscript drafts (.tex files) and research findings, are read from the local file system and sent to an external service (Codex) for analysis. This behavior is consistent with the skill's primary purpose.
[PROMPT_INJECTION]: An indirect prompt injection surface exists as the skill processes user-supplied manuscripts. Malicious instructions embedded within a paper draft could potentially influence the external critic's evaluation, although the impact is mitigated by the advisory nature of the output.

external-review