The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill is designed to manage and retrieve CI/CD secrets. Commands such as glab variable get <key> and glab variable export provide the agent with direct access to raw secret values and private configuration data.
[COMMAND_EXECUTION] (MEDIUM): The skill's documentation explicitly suggests using eval $(glab variable export). This is a dangerous pattern that executes the output of a command which may contain externally-controlled data (variable values), potentially leading to arbitrary command execution if a variable name or value is maliciously crafted.
[DATA_EXFILTRATION] (MEDIUM): The export and list functionalities allow for bulk extraction of sensitive repository metadata and secrets. If the agent's behavior is redirected via prompt injection, these tools facilitate the rapid exfiltration of an organization's entire CI/CD secret store.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill processes untrusted data from an external source (GitLab CI variables).
Ingestion points: CI/CD variable keys and values are ingested through glab variable list and glab variable get.
Boundary markers: There are no boundary markers or instructions to ignore embedded commands within the variable values.
Capability inventory: The skill possesses powerful capabilities including Bash execution and full CRUD (Create, Read, Update, Delete) access to GitLab variables.
Sanitization: No sanitization or validation of variable content is performed before the data is returned to the agent context.

gitlab-variable