The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains no instructions designed to override the agent's safety protocols or disregard previous guidelines. It focuses purely on analytical stress-testing of provided plans.
[DATA_EXFILTRATION]: The skill utilizes tools such as gh, opensrc, and web search to collect data relevant to the plan being analyzed. These tools are used within their intended scope for information retrieval and do not appear to access sensitive system files or exfiltrate data to unauthorized third-party domains.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external project plans as its primary data source. While there are no specific sanitization steps or boundary markers described, the risk is low as the skill's output is limited to generating a 'Challenge Report' rather than performing high-privilege operations.
[COMMAND_EXECUTION]: The skill references standard CLI tools like gh for repository interaction. There is no evidence of arbitrary shell command injection, privilege escalation (e.g., sudo), or persistence mechanisms.

pre-mortem