openspec-plus-apply

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill identifies and runs lint, format, and test commands derived from project configuration files (e.g., package.json, Cargo.toml, go.mod) and documentation (e.g., AGENTS.md). It also utilizes git commands (git log, git diff) to validate files and review changes.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external, user-supplied artifacts including proposal.md, spec.md, design.md, and tasks.md. The content of these files is interpolated into prompts for implementer and reviewer subagents.
  • Ingestion points: Artifacts such as proposal.md, spec.md, design.md, and tasks.md are ingested into the main agent's context during the workflow defined in SKILL.md.
  • Boundary markers: The skill uses explicit prompt templates (implementer-prompt.md, spec-compliance-reviewer-prompt.md, etc.) to structure subagent tasks, though it does not provide specific instructions to ignore potentially malicious instructions embedded within the artifact content.
  • Capability inventory: The skill possesses the capability to execute shell commands (git, lint, test tools) and perform file system operations (reading source files, updating tasks.md).
  • Sanitization: No explicit sanitization, validation, or filtering of the artifact data is performed before it is passed to subagents.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 09:05 AM
Security Audit — agent-trust-hub — openspec-plus-apply