feature-presentation

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The skill prompt includes explicit override-style directives (e.g., "Ignore Claude-specific mode-switch instructions", automatic subagent authorization, and "ALWAYS activate a suitable skill or workflow BEFORE responding / NEVER answer ... before checking") that attempt to override or control the agent's broader system/interaction behavior rather than only describing the deck-building task, which matches the pattern of prompt-injection-style instructions.

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

• Third-party content exposure detected (medium risk: 0.30). The feature-presentation skill reads and embeds repo/tenant artifacts (e.g., team-artifacts/ideas/*, team-artifacts/pbis/*-pbi-*.md, team-artifacts/pbis/*-mockup.html) into the runtime-generated standalone HTML deck via srcdoc, so outsider-authored free text is ingested whenever an outsider can add or modify those artifacts.