arkcli-profile
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
arkclibinary to perform various management tasks such as creating, deleting, and switching profiles, as well as refreshing API keys. These actions are restricted to the local environment and the vendor's official control plane. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data such as profile names, project identifiers, and resource IDs which are passed to CLI commands. The skill implements several defensive measures to mitigate injection risks.
- Ingestion points: User input is accepted for profile names, modality IDs, and project names across multiple files including
SKILL.mdandreferences/arkcli-profile-set-default.md. - Boundary markers: The instructions mandate that the agent must first run
arkcli profile list --format jsonto verify targets before taking action, rather than relying on user-provided strings directly. It also requires the agent to repeat the planned impact and obtain explicit user confirmation for write operations. - Capability inventory: The skill can create, delete, and modify local configuration files (
config.yaml,.env) and fetch API keys from the vendor's control plane. - Sanitization: The skill emphasizes using structured JSON output from the CLI for all checks and enforces exact matching for profile names to prevent command confusion or unauthorized changes.
Audit Metadata