openspec-propose
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
openspecCLI tool through Bash commands to manage project changes and artifacts. It includes commands for creating new changes (openspec new change), checking status (openspec status), and retrieving artifact-specific instructions (openspec instructions). The instructions utilize double-quoting for variable interpolation (e.g.,"<name>"), which helps mitigate command injection risks from user-provided inputs. - [INDIRECT_PROMPT_INJECTION]: The workflow involves reading previously generated artifact files from the local filesystem to provide context for subsequent steps. This ingestion of external data is protected by explicit 'Planning boundary' instructions that forbid the agent from implementing code or editing project files during the planning phase. The skill also requires the use of structured templates for all outputs and instructs the agent to treat external data as constraints rather than content to be copied directly.
Audit Metadata