openspec-archive-change
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external command outputs that provide behavioral guidance.
- Ingestion points: JSON payloads returned by
openspec instructions archivein Step 1 andopenspec instructions specsin Step 4, containing fields likecontext,operationGuidance, andrules. - Boundary markers: The skill includes explicit instructions to keep external inputs separate from built-in steps, user choices, and CLI contracts. It mandates reporting conflicts and prohibits inferring paths or flags from this data.
- Capability inventory: The skill utilizes
Bash(openspec:*)for CLI operations,mkdirfor directory creation, andmvfor moving change roots. - Sanitization: The instructions require the agent to consider guidance as optional and additive, ensuring it does not override primary control logic or existing safety checks.
Audit Metadata