openspec-propose
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes several openspec CLI commands via Bash, such as openspec new change, openspec status, and openspec instructions, to manage the planning lifecycle and state. These commands are restricted to the authorized openspec binary.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from existing project files and CLI-generated instructions to generate new content. This is a core feature of the planning workflow.\n
- Ingestion points: Reads project artifact files (markdown) and JSON output from OpenSpec CLI commands.\n
- Boundary markers: Instructions guide the agent to use provided context as constraints but do not define explicit delimiters for untrusted content.\n
- Capability inventory: Includes executing openspec commands and writing generated files to the filesystem.\n
- Sanitization: Relies on the agent's internal processing and the structure of the OpenSpec CLI templates.
Audit Metadata