The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it incorporates external repository data into the agent's context. • Ingestion points: Output from gh issue list is injected into the context in SKILL.md. • Boundary markers: The skill lacks delimiters or specific instructions to isolate or ignore embedded commands in the issue data. • Capability inventory: The skill can execute gh issue create, gh label create, and git commands as defined in SKILL.md and references/decision-logic.md. • Sanitization: No sanitization or validation of the external content is performed.
[COMMAND_EXECUTION]: Dynamic context injection is used in SKILL.md to execute git status, git branch, gh issue list, and gh auth status when the skill is loaded. These commands provide necessary environmental information and are used in a manner consistent with the skill's stated purpose without accessing sensitive files or unauthorized networks.

create-issues