The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes shell commands using the GitHub CLI (gh issue create, gh label create, gh sub-issue add) by interpolating strings directly from user input and document content.
Evidence: gh issue create --title "<title>" --label "<labels>" --body "<body>" in SKILL.md (Standalone Mode) and similar patterns in references/usdm-issue-mapping.md.
Risk: Without explicit sanitization of shell metacharacters in titles or bodies, there is a risk of command injection if the environment does not provide automatic escaping.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and parse external Markdown files (USDM documents) which may contain instructions that could influence the agent's behavior.
Ingestion points: SKILL.md (Mode 2) instructs the agent to read a Markdown file at a user-provided path.
Boundary markers: Absent. The instructions do not provide delimiters or specific warnings to ignore instructions embedded within the document content.
Capability inventory: The agent can create issues, create labels, and modify repository hierarchies using the gh tool.
Sanitization: Absent. Content extracted from the document (Reason, Description, specifications) is directly interpolated into issue bodies and shell commands.

issue-create