The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from two primary sources: user-provided descriptions of bugs or features, and existing issue bodies retrieved from remote repositories during duplicate resolution.
Ingestion points: User input sentences defined in the Trigger section of SKILL.md and existing issue descriptions fetched via gh issue view or glab issue view in references/create.md §2b.
Boundary markers: The skill uses structured Markdown templates (assets/bug-report.md, assets/feature-request.md) to separate different types of information, which provides some structural delimitation.
Capability inventory: The skill utilizes the gh and glab CLI tools to perform network write operations, including creating new issues and updating existing ones (references/host-cli.md).
Sanitization: The skill demonstrates security awareness by explicitly recommending the use of body files (-F) for GitHub to avoid shell quoting and injection issues associated with inline strings. However, for GitLab, it uses command substitution $(cat body.md) which could be sensitive to certain characters if not handled by the underlying execution environment.

issue