The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the GitHub CLI (gh) to interact with repository issues. It uses this to check authentication status, verify repository settings, search for duplicate bug reports, and create new issues. These operations are restricted to specific, pre-defined vendor repositories.
[COMMAND_EXECUTION]: In the SKILL.md instructions, shell commands and Node.js one-liners are used to dynamically discover the plugin's installation path. This mechanism allows the agent to locate the supporting report-bug.cjs script regardless of where the plugin is installed.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied bug descriptions and system environment data. It mitigates injection risks through several layers of defense:
Ingestion points: User input from the conversation and diagnostic data from the local environment.
Boundary markers: The final report is structured using Markdown headers and code blocks.
Capability inventory: The skill has the ability to write temporary JSON files (fs.writeFileSync) and execute the GitHub CLI (gh).
Sanitization: The report-bug.cjs script implements an extensive redaction engine that uses regular expressions to strip out API keys, tokens, private keys, local usernames, and absolute file paths from the report before submission. Furthermore, a mandatory manual review step ensures the user sees the final, redacted draft before it is sent to GitHub.

report-videocut-bug