The Agent Skills Directory

Command Execution: The skill uses the GitHub CLI (gh issue create) to submit feedback. This involves executing shell commands to interact with external services. The skill mitigates risks by requiring explicit user confirmation of the generated title and body before submission.
Data Handling and Redaction: The skill processes user prompts and system logs to generate reports. It includes specific instructions to redact sensitive information such as SA passwords, registry credentials, and access tokens before the data is transmitted. This is a positive security practice for handling potentially sensitive information.
Indirect Prompt Injection Surface: The skill ingests untrusted data from user prompts and execution logs to build issue reports.
Ingestion points: User prompts, error messages, and logs are gathered in SKILL.md (Step 2).
Boundary markers: The skill requires showing the user the full report for review before submission (Step 4).
Capability inventory: Uses gh issue create to perform network operations via the CLI.
Sanitization: Step 3 (Redaction) defines a list of sensitive patterns to be stripped from the content.

azuresql-db-feedback