sdd-init

Pass

Audited by Gen Agent Trust Hub on Mar 16, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. It reads and parses untrusted files from the project root and user-level AI tool directories to generate configuration and registry artifacts.
  • Ingestion points: The skill reads content from project-level files such as package.json, go.mod, pyproject.toml, agents.md, CLAUDE.md, and .cursorrules. It also scans and reads SKILL.md files from user-level directories like ~/.claude/skills/ and ~/.cursor/skills/ during the registry build phase (Step 4).
  • Boundary markers: Absent. The skill does not apply delimiters or provide instructions to the agent to ignore potentially malicious embedded commands within the files it ingests.
  • Capability inventory: The skill has the capability to write to the local file system (creating the openspec/ directory structure and .atl/skill-registry.md) and persists data to the agent's memory backend via the mem_save function (Step 5).
  • Sanitization: Absent. No explicit sanitization, validation, or escaping of the content read from external files is performed before it is interpolated into the generated SDD configuration or stored in memory.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 16, 2026, 01:57 PM