diff --git a/prompts/README.md b/prompts/README.md
new file mode 100644
index 0000000..a77b30d
--- /dev/null
+++ b/prompts/README.md
@@ -0,0 +1 @@
+Community-contributed prompts for PromptLayer.
diff --git a/prompts/prompt-evaluation-ambiguous-inputs.md b/prompts/prompt-evaluation-ambiguous-inputs.md
new file mode 100644
index 0000000..ec60e80
--- /dev/null
+++ b/prompts/prompt-evaluation-ambiguous-inputs.md
@@ -0,0 +1,30 @@
+# Prompt Evaluation for Ambiguous User Inputs
+
+## Purpose
+This prompt is designed to evaluate how well a system prompt handles
+ambiguous or underspecified user inputs, which are common in real-world
+usage.
+
+## Prompt
+"You are an assistant that must identify ambiguity in user queries.
+If the user's request lacks sufficient detail, clearly state what is
+missing and ask focused follow-up questions before attempting an answer."
+
+## When this works well
+- Early-stage product prototypes
+- User-facing AI systems with diverse audiences
+- Scenarios where incorrect assumptions are costly
+
+## Common failure modes
+- The model attempts to guess user intent instead of clarifying
+- Over-asking questions, leading to poor user experience
+- Asking generic clarifications instead of targeted ones
+
+## Strategic considerations
+This prompt should be paired with:
+- A maximum clarification threshold
+- Clear rules for when to proceed vs stop
+- Human review for high-risk decisions
+
+Prompt evaluation should focus on user trust and clarity, not just
+response completeness.