The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the acquisition and installation of the GitLab Duo CLI executable using the glab duo cli --install and glab duo cli update commands. These operations can be configured to run non-interactively using the --yes flag or through global configuration settings like duo_cli_auto_download.
[COMMAND_EXECUTION]: The skill executes the glab CLI to interface with the GitLab Duo Agent Platform. This includes the ability to perform autonomous actions on a codebase via the run --goal subcommand, which takes natural language instructions to modify code or fix tests.
[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow where an AI agent processes codebase data and user-defined goals, creating a vulnerability surface for indirect injection.
Ingestion points: Data is ingested from the local project working tree and user input via the --goal parameter as documented in SKILL.md and references/commands.md.
Boundary markers: The skill does not specify the use of delimiters or specific instructions to ignore embedded commands within the files being processed.
Capability inventory: The glab tool has permissions to read repository content and interact with authenticated GitLab service endpoints.
Sanitization: No explicit sanitization or filtering is performed on the codebase data, though the skill provides best-practice guidance such as using isolated working trees and performing independent reviews of agent-generated changes.

glab-duo