skills

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

• Third-party content exposure detected (medium risk: 0.30). Skill.md defines that user-supplied free-form text in $ARGUMENTS selects the mode/targets and then dispatches into create/improve/review flows whose subagents read project-local skill directories (e.g., .claude/skills/*/SKILL.md, references/*.md, scripts/, etc.), meaning outsider text can drive which already-present outsider-authored artifacts are ingested via the repository file-reading/search mechanisms used by the skill-creator/review agents.