The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation includes examples of git and pytest commands within plan templates. These are standard development operations intended to guide the implementor and do not represent unauthorized command execution.
[DATA_EXFILTRATION]: There are no indicators of unauthorized network activity. The skill prescribes saving plan documents to a local directory (docs/quirk/plans/), which is a standard practice for project documentation.
[PROMPT_INJECTION]: The instructions follow a structured workflow without any attempts to override model safety filters, ignore prior instructions, or extract system prompts.
[REMOTE_CODE_EXECUTION]: The skill enforces a strict 'No-Code Rule,' explicitly prohibiting the inclusion of runnable implementation code or full test bodies in plans. This policy significantly reduces the risk of generating or executing malicious code snippets.

writing-plans