workspace-orchestration

Pass

Audited by Gen Agent Trust Hub on Oct 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill describes the use of a local CLI tool named 'docko' for orchestrating workspace slots and managing sessions via environment variables like DOCKO_SESSION_ID.
  • [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to execute a 'suggested_command' if it receives an 'AMBIGUOUS_SESSION' error from the docko tool. This introduces a risk where tool-generated content can directly control agent behavior. 1. Ingestion points: Error payloads from the docko CLI tool (SKILL.md). 2. Boundary markers: None present to delineate or validate the suggested command. 3. Capability inventory: Shell command execution capabilities (invoking docko and suggested commands). 4. Sanitization: No sanitization or verification steps are described for the dynamically provided command.
  • [DYNAMIC_EXECUTION]: The instruction to run a command provided by the tool at runtime (the suggested_command payload) constitutes dynamic execution of externally generated instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 5, 2026, 11:44 PM