openspec
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCEEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external project artifacts (Markdown files), creating a surface where malicious instructions in those files could potentially influence agent actions during implementation.
- Ingestion points: Project artifacts located in the change directory and
specs/folder. - Boundary markers: No specific delimiters or safety instructions are defined for processing untrusted artifact content.
- Capability inventory: The skill uses the
openspecCLI to manage files and state, and instructs the agent to perform implementation tasks based on artifact content. - Sanitization: No explicit sanitization of artifact content is mentioned.
- [PERSISTENCE]: The skill describes the
openspec initcommand, which includes a feature to install shell completions. This modifies shell profiles (e.g.,.bashrc, PowerShell profiles). The skill explicitly notes this as an 'opt-in' configuration requiring user consent. - [EXTERNAL_DOWNLOADS]: The skill references documentation and installation links on GitHub and NPM, which are well-known and established service providers.
Audit Metadata