The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses mandatory language to command the agent to "Do not use your plan mode," which is an attempt to override the agent's standard internal reasoning and planning workflows.
[PROMPT_INJECTION]: The skill points to external files (../alignfirst/SKILL.md and references/review-protocol.md) as the authoritative source for its execution protocol. This creates a surface for indirect prompt injection where external content, which could be modified independently of this skill, dictates the agent's actions.

alreview