The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from GitHub issue titles, PR bodies, and local plan files to drive its logic (e.g., determining dependencies, milestones, and phases).
Ingestion points: Data enters the context via gh issue view, gh pr list, and the Read tool on files in docs/plans/.
Boundary markers: Absent. The instructions do not specify using delimiters or instructions to ignore embedded prompts when parsing these external strings.
Capability inventory: The skill can execute shell commands (gh, git), edit files, and invoke other skills.
Sanitization: The skill uses --body-file for multi-line inputs to avoid certain shell injection vectors, but does not specify escaping for interpolated variables like titles or search queries.
[COMMAND_EXECUTION]: The skill constructs shell commands using the gh CLI where variables (like issue numbers, titles, or search terms) are interpolated directly into shell strings.
Evidence: Commands like gh pr list --search "<n>" in modes/close.md or gh search issues ... --match title "<candidate title>" in modes/create.md depend on the agent correctly handling potentially malicious characters in those variables.

issuekit