Improve PR failure issue title specificity for engine HTTP 429 rate limits

actions/setup/js/handle_agent_failure.cjs

@@ -253,6 +253,7 @@ function buildFailureMatchCategories(options) {
   if (options.modelNotSupportedError) categories.push("model_not_supported_error");
   if (options.http400ResponseError) categories.push("http_400_response_error");
   if (options.aiCreditsRateLimitError) categories.push("ai_credits_rate_limit_error");
+  if (options.hasEngineRateLimit429) categories.push("engine_rate_limit_429");
   if (options.unknownModelAICredits) categories.push("unknown_model_ai_credits");
   if (options.missingModelPricingError) categories.push("missing_model_pricing");
   if (options.maxAICreditsExceeded) categories.push("max_ai_credits_exceeded");
@@ -291,6 +292,7 @@ function buildFailureMatchCategories(options) {
  * @param {boolean} options.hasStaleLockFileFailed
  * @param {boolean} options.hasDailyAICExceeded
  * @param {boolean} options.aiCreditsRateLimitError
+ * @param {boolean} options.hasEngineRateLimit429
  * @param {boolean} options.maxAICreditsExceeded
  * @param {boolean} options.hasAssignmentErrors
  * @param {boolean} options.http400ResponseError
@@ -304,6 +306,7 @@ function buildFailureIssueTitle(options) {
   if (options.hasDailyAICExceeded) return `[aw] ${workflowName} exceeded daily AI credits budget`;
   if (options.maxAICreditsExceeded) return `[aw] ${workflowName} exceeded max AI credits`;
   if (options.aiCreditsRateLimitError) return `[aw] ${workflowName} hit AI credits rate limit`;
+  if (options.hasEngineRateLimit429) return `[aw] ${workflowName} hit engine rate limit (HTTP 429)`;
   // Missing model pricing is surfaced by the proxy as HTTP 400, so prefer the
   // specialized title before falling back to the generic transport-level error.
   if (options.missingModelPricingError) {
@@ -2614,6 +2617,33 @@ function detectAWFFirewallStartupFailureFromLog() {
   }
 }
 
+/**
+ * Detect whether the agent failure was caused by engine HTTP 429/rate limiting.
+ * Checks agent-stdio.log first, then falls back to OTLP mirror payloads.
+ * @returns {boolean}
+ */
+function detectEngineRateLimit429Failure() {
+  const agentOutputFile = process.env.GH_AW_AGENT_OUTPUT;
+  const stdioLogPath = agentOutputFile ? path.join(path.dirname(agentOutputFile), "agent-stdio.log") : "/tmp/gh-aw/agent-stdio.log";
+  try {
+    if (fs.existsSync(stdioLogPath)) {
+      const logContent = fs.readFileSync(stdioLogPath, "utf8");
+      // If the agent completed successfully (terminal_reason: "completed"), the failure
+      // was caused by something other than the agent itself. Suppress the 429 signal to
+      // avoid giving a rate-limit title to an unrelated post-processing failure.
+      if (/"terminal_reason"[ ]?:[ ]?"completed"/.test(logContent)) {
+        return false;
+      }
+      if (hasEngineRateLimit429Signal(logContent)) {
+        return true;
+      }
+    }
+  } catch {
+    // Ignore read errors and continue with OTLP mirror fallback.
+  }
+  return hasEngineRateLimit429InOTELMirror();
+}
+
 /**
  * Extract terminal error messages from agent-stdio.log to surface engine failures.
  * First tries to match known error patterns (ERROR:, Error:, Fatal:, panic:, Reconnecting...).
@@ -3389,6 +3419,7 @@ async function main() {
     if (hasToolDenialsExceeded) {
       core.info(`Detected ${toolDenialsExceededEvents.length} guard.tool_denials_exceeded event(s) from Copilot SDK events.jsonl`);
     }
+    const hasEngineRateLimit429 = agentConclusion === "failure" && !maxAICreditsExceeded && !aiCreditsRateLimitError && detectEngineRateLimit429Failure();
 
     // Detect cache-miss misconfiguration: the agent reported a missing_data with reason
     // "cache_memory_miss" after a cache restore matched. This indicates the prompt
@@ -3537,6 +3568,7 @@ async function main() {
       hasStaleLockFileFailed,
       hasDailyAICExceeded,
       aiCreditsRateLimitError,
+      hasEngineRateLimit429,
       maxAICreditsExceeded,
       hasAssignmentErrors,
       http400ResponseError,
@@ -3566,6 +3598,7 @@ async function main() {
       modelNotSupportedError,
       http400ResponseError,
       aiCreditsRateLimitError,
+      hasEngineRateLimit429,
       unknownModelAICredits,
       missingModelPricingError,
       maxAICreditsExceeded,
@@ -4141,6 +4174,7 @@ module.exports = {
   hasEngineMaxRunsExceededSignal,
   hasEngineRateLimit429Signal,
   hasEngineRateLimit429InOTELMirror,
+  detectEngineRateLimit429Failure,
   buildEngineMaxRunsExceededContext,
   buildEngineRateLimit429Context,
   hasEngineMaxCacheMissesExceededSignal,

actions/setup/js/handle_agent_failure.test.cjs

@@ -97,6 +97,7 @@ describe("handle_agent_failure", () => {
       hasStaleLockFileFailed: false,
       hasDailyAICExceeded: false,
       aiCreditsRateLimitError: false,
+      hasEngineRateLimit429: false,
       maxAICreditsExceeded: false,
       hasAssignmentErrors: false,
       http400ResponseError: false,
@@ -109,6 +110,7 @@ describe("handle_agent_failure", () => {
       { flag: "hasDailyAICExceeded", expected: "[aw] Test Workflow exceeded daily AI credits budget" },
       { flag: "maxAICreditsExceeded", expected: "[aw] Test Workflow exceeded max AI credits" },
       { flag: "aiCreditsRateLimitError", expected: "[aw] Test Workflow hit AI credits rate limit" },
+      { flag: "hasEngineRateLimit429", expected: "[aw] Test Workflow hit engine rate limit (HTTP 429)" },
       { flag: "http400ResponseError", expected: "[aw] Test Workflow hit HTTP 400 bad request" },
       { flag: "unknownModelAICredits", expected: "[aw] Test Workflow has unknown model pricing" },
       { flag: "hasAppTokenMintingFailed", expected: "[aw] Test Workflow failed to mint GitHub App token" },
@@ -155,6 +157,10 @@ describe("handle_agent_failure", () => {
         "[aw] Test Workflow has no AI credits pricing for model (claude-opus-5)"
       );
     });
+
+    it("aiCreditsRateLimitError takes precedence over hasEngineRateLimit429 when both are true", () => {
+      expect(buildFailureIssueTitle({ ...baseOptions, aiCreditsRateLimitError: true, hasEngineRateLimit429: true })).toBe("[aw] Test Workflow hit AI credits rate limit");
+    });
   });
 
   describe("detection caution placement in main()", () => {
@@ -2945,6 +2951,62 @@ describe("handle_agent_failure", () => {
     });
   });
 
+  // ──────────────────────────────────────────────────────
+  // detectEngineRateLimit429Failure
+  // ──────────────────────────────────────────────────────
+
+  describe("detectEngineRateLimit429Failure", () => {
+    let detectEngineRateLimit429Failure;
+    const fs = require("fs");
+    const path = require("path");
+    const os = require("os");
+
+    /** @type {string} */
+    let tmpDir;
+    /** @type {string} */
+    let stdioLogPath;
+
+    beforeEach(() => {
+      vi.resetModules();
+      tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), "aw-test-detect-429-"));
+      stdioLogPath = path.join(tmpDir, "agent-stdio.log");
+      process.env.GH_AW_AGENT_OUTPUT = path.join(tmpDir, "agent_output.json");
+      ({ detectEngineRateLimit429Failure } = require("./handle_agent_failure.cjs"));
+    });
+
+    afterEach(() => {
+      delete process.env.GH_AW_AGENT_OUTPUT;
+      delete process.env.GH_AW_OTEL_JSONL_PATH;
+      if (fs.existsSync(tmpDir)) {
+        fs.rmSync(tmpDir, { recursive: true, force: true });
+      }
+    });
+
+    it("returns true when stdio log contains a 429 rate-limit signal", () => {
+      fs.writeFileSync(stdioLogPath, "Failed to get response from the AI model; retried 5 times. Last error: CAPIError: 429 429 Sorry, you've exceeded your rate limit.\n");
+      expect(detectEngineRateLimit429Failure()).toBe(true);
+    });
+
+    it("returns false when stdio log contains terminal_reason: completed even with a 429 signal", () => {
+      fs.writeFileSync(stdioLogPath, 'Failed to get response; CAPIError: 429 rate limit\n{"type":"result","subtype":"success","terminal_reason":"completed","num_turns":10}\n');
+      expect(detectEngineRateLimit429Failure()).toBe(false);
+    });
+
+    it("returns false when stdio log contains terminal_reason: completed and no 429 signal", () => {
+      fs.writeFileSync(stdioLogPath, '{"type":"result","subtype":"success","terminal_reason":"completed","num_turns":5}\n');
+      expect(detectEngineRateLimit429Failure()).toBe(false);
+    });
+
+    it("returns false when stdio log has no 429 signal and OTLP mirror is absent", () => {
+      fs.writeFileSync(stdioLogPath, "Agent exited normally.\n");
+      expect(detectEngineRateLimit429Failure()).toBe(false);
+    });
+
+    it("returns false when log file does not exist and OTLP mirror is absent", () => {
+      expect(detectEngineRateLimit429Failure()).toBe(false);
+    });
+  });
+
   // ──────────────────────────────────────────────────────
   // hasEngineMaxCacheMissesExceededSignal
   // ──────────────────────────────────────────────────────
@@ -5175,6 +5237,23 @@ describe("handle_agent_failure", () => {
       });
       expect(categories).not.toContain("missing_model_pricing");
     });
+
+    it("returns engine_rate_limit_429 category when hasEngineRateLimit429 is true", () => {
+      const categories = buildFailureMatchCategories({
+        agentConclusion: "failure",
+        hasEngineRateLimit429: true,
+      });
+      expect(categories).toContain("engine_rate_limit_429");
+      expect(categories).not.toContain("agent_failure");
+    });
+
+    it("does not return engine_rate_limit_429 category when hasEngineRateLimit429 is false", () => {
+      const categories = buildFailureMatchCategories({
+        agentConclusion: "failure",
+        hasEngineRateLimit429: false,
+      });
+      expect(categories).not.toContain("engine_rate_limit_429");
+    });
   });
 
   describe("failure categories filter behavior", () => {