The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute gh and glab CLI commands for issue management. Security is maintained by a restricted allowed-tools list that limits execution to specific, pre-defined subcommands, effectively reducing the command execution attack surface.
[DATA_EXPOSURE]: The skill ingests session context and repository data to help draft issues. It includes a specific security guardrail that scans the resulting draft for sensitive patterns like API keys, tokens, and .env content, excluding them to prevent data exfiltration.
[INDIRECT_PROMPT_INJECTION]: Because the skill processes issue bodies and comments from external sources, it is potentially exposed to indirect prompt injection. It mitigates this by implementing an 'Author authority' framework, which verifies the identity and permissions (e.g., maintainer status) of content authors before treating their text as instructions.
[REMOTE_CODE_EXECUTION]: The skill invokes a local shell script (templates/resolve-config.sh) to process its configuration. This script is part of the skill's own package and is used for internal logic rather than executing arbitrary remote code.

issue