openspec-plus-proposal

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from project-level instruction files (like AGENTS.md and CLAUDE.md) and existing proposal artifacts to inform the discovery and writing phases.
  • Ingestion points: Reads local documentation and existing change files from the project root and 'openspec/changes/' directory.
  • Boundary markers: The skill does not implement specific delimiters or warnings for the agent to ignore instructions embedded within the ingested files.
  • Capability inventory: The skill can execute the 'openspec' CLI, write to local state files (timestamp and VERSION), and dispatch subagents for review.
  • Sanitization: No explicit logic is provided to sanitize or filter the content read from external project files.
  • [EXTERNAL_DOWNLOADS]: Fetches version metadata from the author's official GitHub repository ('api.github.com/repos/sudokar/openspec-plus/releases/latest') to perform a session-based auto-update check.
  • [COMMAND_EXECUTION]: Executes the 'openspec' command-line interface to resolve templates and per-section guidance in JSON format based on the current project context.
  • [DYNAMIC_EXECUTION]: Employs platform-native multi-agent tools (such as 'run_subagent' or 'spawn_agent') to dispatch a reviewer subagent that performs a single-shot compliance check on the generated proposal document.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 01:15 AM
Security Audit — agent-trust-hub — openspec-plus-proposal