The Agent Skills Directory

[SAFE]: The skill instructions are limited to text processing and issue drafting, with explicit rules against automated submission, commenting, or modification of remote issue trackers.
[SAFE]: The frontmatter configuration disable-model-invocation: true scopes the skill to a restricted environment where the agent cannot invoke additional tools, mitigating potential injection risks.
[SAFE]: Local file access is strictly bounded to reading repository documentation for context and saving drafts only when a prior user agreement exists.
[SAFE]: References to the simple-english style component refer to modular instructions already present in the environment rather than external code downloads or installations.

issue-writing