auto-skill

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

• Third-party content exposure detected (medium risk: 0.30). Skill Seekers in the required workflow can be invoked with scraping/fetching from outsider-provided sources (docs/GitHub/PDF) via skill-seekers.sh, which then ingests the resulting free text into the generated SKILL.md/references/ for the model to use.

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

• Potentially malicious external URL detected (high risk: 0.90). scripts/skill-seekers-update.sh downloads and extracts a remote GitHub archive at runtime using the URL "https://codeload.github.com/${repo}/tar.gz/${ref}", and that linked Skill_Seekers source is a required tool which is later executed via scripts/skill-seekers.sh (python -m skill_seekers...), so remote code can be fetched and run affecting agent behavior.