The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes explicit defensive instructions in 'Hard Rule 4' to mitigate indirect prompt injection. It commands the agent to treat all ingested repository content as inert data and to specifically ignore and flag common injection markers like 'ignore previous instructions'. This demonstrates a security-conscious design.
[COMMAND_EXECUTION]: The skill uses the local command git rev-parse --short HEAD to version-stamp generated implementation plans. Additionally, it defines an execute workflow variant that dispatches subagents to apply changes in isolated worktrees; while this involves code modification, the skill provides a restricted template (PLAN-TEMPLATE.md) and requires subagents to adhere to the same safety posture.
[SAFE]: The external reference to 'emilkowal.ski' in AUDIT.md is a legitimate vendor resource belonging to the skill author, used solely for establishing design philosophy and animation standards.

improve-animations