The Agent Skills Directory

[Command Execution]: The skill executes local shell commands (claude --version, pac help, and OS detection commands) to gather environment information for the bug report.
Evidence: Mentioned in report-issue-workflow.md under Phase 2, Call 3, where commands are run to populate environment context.
Context: The design implements a safeguard by explicitly asking the user via AskUserQuestion before performing any collection, mitigating unauthorized automated execution.
[Interaction with External Platform (GitHub CLI)]: The skill invokes the gh tool via Bash to submit data to a public repository.
Evidence: Indicated in Phase 4 of report-issue-workflow.md where gh issue create is used.
Context: This is aligned with the skill's core purpose. The workflow enforces user review by outputting a full issue text preview and requiring approval before submission, helping prevent accidental disclosure of private information.
[Indirect Prompt Injection Surface]: The skill handles user-supplied external arguments and descriptions, which are later rendered and included in shell commands or text outputs.
Evidence: $ARGUMENTS and user responses collected via AskUserQuestion flow into Phase 3 (Preview) and Phase 4 (Issue Creation).
Boundary Markers: Absent from direct command invocation, but mitigated because data is passed into a gh heredoc block and explicitly displayed to the user beforehand.
Capability Inventory: Uses Bash for command execution and GitHub interactions.
Sanitization: Does not explicitly detail syntax escaping for shell arguments, but mandates user review steps prior to execution.

report-issue