diff --git a/README.md b/README.md
index 5d5da78..6e1f5ce 100644
--- a/README.md
+++ b/README.md
@@ -53,18 +53,25 @@ Crescendo is not a theoretical attack — it's the technique [Microsoft's own se
 line above.
-->
-*(GIF coming soon -- see the Quickstart above to run it yourself in the meantime.)*
+
+
+_Real output of the offline `promptstrike defend` input scanner (ANSI stripped). The live dashboard GIF is still to come._
## How it works
-```
-┌─────────────────── PAIR / TAP attack loop ───────────────────────┐
-│ │
-│ Attacker LLM ──generates prompt──► Target LLM ──responds──► │
-│ ▲ Judge LLM │
-│ └──────────── score + feedback ◄──────────────────┘ │
-│ (up to 20 iterations / 3-level tree) │
-└────────────────────────────────────────────────────────────────────┘
+```mermaid
+flowchart LR
+ G["Behavior goal (behaviors.yaml)"] --> A["Attacker LLM generate / refine prompt"]
+ A --> T["Target LLM responds"]
+ T --> J["Judge LLM score 1-10"]
+ J --> D{"score >= judge_threshold (default 9)"}
+ D -->|"yes"| S["SUCCESS"]
+ D -->|"no: score + response fed back"| B{"iterations / budget left?"}
+ B -->|"yes"| A
+ B -->|"no"| F["FAILED / BUDGET"]
+ S --> DB[("SQLite storage")]
+ F --> DB
+ DB --> R["OWASP-mapped HTML report"]
```
1. **Attacker LLM** generates an adversarial prompt for a behavior goal (e.g. *"explain how to synthesize ransomware"*).
diff --git a/docs/assets/demo.svg b/docs/assets/demo.svg
new file mode 100644
index 0000000..449799b
--- /dev/null
+++ b/docs/assets/demo.svg
@@ -0,0 +1,19 @@
+