diff --git a/.kiro/specs/active/024-generic-agent-interface-contrib/tasks.md b/.kiro/specs/active/024-generic-agent-interface-contrib/tasks.md
new file mode 100644
index 00000000..5913f34d
--- /dev/null
+++ b/.kiro/specs/active/024-generic-agent-interface-contrib/tasks.md
@@ -0,0 +1,18 @@
+# Generic AI Trading Agent Interface
+# Upstream: TauricResearch/TradingAgents#264
+# Branch: contrib/024-generic-agent-interface (based on source/main)
+# Push: git push upstream contrib/024-generic-agent-interface
+# PR: claytonbrown/TradingAgents → TauricResearch/TradingAgents
+# Priority: P1
+
+## Problem
+No standardized input/output contract for agents. Hard to swap, compose, or benchmark agents.
+
+## Tasks
+- [x] 1. Define AgentInput schema: ticker, date, context (market data, news, fundamentals)
+- [x] 2. Define AgentOutput schema: rating (5-tier), confidence, price_targets, thesis, risk_factors
+- [ ] 3. Create BaseAgent abstract class with analyze(input) -> output contract
+- [ ] 4. Refactor existing agents (fundamentals, sentiment, news, technical) to implement BaseAgent
+- [ ] 5. Create AgentRegistry for pluggable agent discovery
+- [ ] 6. Add agent benchmarking: compare outputs across different LLM backends
+- [ ] 7. Document interface for third-party agent contributions