video-analyser

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly downloads arbitrary direct video URLs with curl ("Download: curl -fL -o "$WORK_DIR/input.mp4" "$INPUT"") and pulls attachment URLs from Linear issues via the Linear MCP (Step 1), then runs OCR and/or audio transcription and injects that untrusted, user-generated text/audio into the analysis prompt (Steps 6–8), meaning third-party content is fetched and directly interpreted as part of the agent's workflow.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill downloads arbitrary external video URLs at runtime (curl -fL -o "$WORK_DIR/input.mp4" "$INPUT") and Linear attachment URLs (curl -fL -o "$WORK_DIR/input.mp4" "$ATTACHMENT_URL"), then extracts frames/OCR/audio and injects that extracted text/images directly into the assembled analysis prompt, meaning externally hosted content can control model input.