The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interacts with the GitLab API through the official glab command-line interface. This is a standard administrative task for managing repository and group access.
[CREDENTIALS_SAFE]: The documentation identifies that the create command outputs sensitive one-time secrets. It provides specific instructions to avoid leaking these secrets in logs, terminal recordings, or task reports, recommending an operator-controlled handoff to a secret manager.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it retrieves metadata from external GitLab environments, which could theoretically contain malicious instructions in project or token names.
Ingestion points: Metadata is ingested via the glab deploy-token list and glab deploy-token get commands.
Boundary markers: The skill mandates explicit human approval for the host, project, and actor before any read or write operation is performed.
Capability inventory: The skill is capable of executing shell commands via glab.
Sanitization: The instructions note that token values are cleared by the tool before JSON serialization for metadata commands.

glab-deploy-token