migrate-skill-agent

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). Outsider-authored free text from the public/community skills-store Git repository is ingested at runtime via git pull and then copied from files like SKILL.md and agent .md/.toml into the target directories (e.g., $STORE/.claude/skills/$NAME/SKILL.md, $STORE/.codex/agents/$NAME.md), which are then available for the agent/LLM to read.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The script verifies the skills-store remote must match URLs such as https://github.com/Mineru98/skills-store.git (also listed as https://github.com/Mineru98/skills-store, git@github.com:Mineru98/skills-store.git, ssh://git@github.com/Mineru98/skills-store.git) and then runs git pull and copies SKILL.md / agent files from that repo into local .claude/.codex directories at runtime, which lets externally fetched content directly control agent prompts/instructions.