diff --git a/.github/scripts/README.md b/.github/scripts/README.md
index 727787170..d20d271ae 100644
--- a/.github/scripts/README.md
+++ b/.github/scripts/README.md
@@ -16,6 +16,30 @@ between jobs and covered by lightweight unit tests.
   steps can invoke them with a simple `python .github/scripts/<name>.py`
   command.
 
+## Retry Logic for GitHub API Calls
+
+To handle transient failures (rate limits, timeouts, network issues), use the retry helpers from `api-helpers.js`:
+
+```javascript
+const { withBackoff, paginateWithBackoff } = require('./api-helpers');
+
+// For single API calls
+const result = await withBackoff(
+  () => github.rest.pulls.get({ owner, repo, pull_number: 123 }),
+  { core, maxRetries: 3 }
+);
+
+// For paginated calls
+const items = await paginateWithBackoff(
+  github,
+  github.rest.issues.listComments,
+  { owner, repo, issue_number: 123 },
+  { core, maxRetries: 3 }
+);
+```
+
+These helpers automatically retry transient errors (503, 504, rate limits, timeouts) with exponential backoff and jitter.
+
 ## Tests
 
 Minimal Node and Python unit tests live alongside the scripts under
diff --git a/.github/scripts/__tests__/api-helpers.test.js b/.github/scripts/__tests__/api-helpers.test.js
index 27961a1b1..d73b567ff 100644
--- a/.github/scripts/__tests__/api-helpers.test.js
+++ b/.github/scripts/__tests__/api-helpers.test.js
@@ -3,7 +3,8 @@
 const test = require('node:test');
 const assert = require('node:assert/strict');
 
-const { calculateBackoffDelay } = require('../api-helpers');
+const { calculateBackoffDelay } = require('../github_api_retry');
+const { withBackoff, paginateWithBackoff } = require('../api-helpers');
 
 function withStubbedRandom(value, fn) {
   const originalRandom = Math.random;
@@ -28,3 +29,48 @@ test('calculateBackoffDelay applies negative jitter within expected range', () =
     assert.equal(delay, 1500);
   });
 });
+
+test('withBackoff retries transient errors (not just rate limits)', async () => {
+  let attempts = 0;
+  const result = await withBackoff(
+    async () => {
+      attempts += 1;
+      if (attempts < 3) {
+        const error = new Error('Service temporarily unavailable');
+        error.status = 503;
+        throw error;
+      }
+      return { data: 'success' };
+    },
+    { maxRetries: 3 }
+  );
+  
+  assert.equal(attempts, 3);
+  assert.deepEqual(result, { data: 'success' });
+});
+
+test('paginateWithBackoff retries transient errors', async () => {
+  let attempts = 0;
+  const mockGithub = {
+    paginate: async (method, params) => {
+      attempts += 1;
+      if (attempts < 2) {
+        const error = new Error('Network timeout');
+        error.status = 504;
+        throw error;
+      }
+      return [{ id: 1 }, { id: 2 }];
+    },
+  };
+  
+  const mockMethod = 'mockMethod';
+  const result = await paginateWithBackoff(
+    mockGithub,
+    mockMethod,
+    { page: 1 },
+    { maxRetries: 3 }
+  );
+  
+  assert.equal(attempts, 2);
+  assert.deepEqual(result, [{ id: 1 }, { id: 2 }]);
+});
diff --git a/.github/scripts/agents_pr_meta_update_body.js b/.github/scripts/agents_pr_meta_update_body.js
index 6529d40f0..67c189ff7 100644
--- a/.github/scripts/agents_pr_meta_update_body.js
+++ b/.github/scripts/agents_pr_meta_update_body.js
@@ -208,13 +208,12 @@ function upsertBlock(body, marker, replacement) {
 /**
  * Simple retry wrapper with linear backoff for general API errors.
  * 
- * Note: This differs from api-helpers.js `withBackoff` which specifically handles
- * rate limit errors (403/429) with exponential backoff and reset time extraction.
- * This function retries any error type with simple linear delay, suitable for
- * transient network/server errors during PR body updates.
+ * Note: api-helpers.js now has `withBackoff` which handles all transient errors
+ * (not just rate limits) with exponential backoff. This function uses linear delay
+ * which may be preferred for specific use cases in PR body updates.
  * 
  * @param {Function} fn - Async function to retry
- * @param {Object} options - Configuration options
+ * @param {Object} [options] - Retry configuration options
  * @param {number} [options.attempts=3] - Number of attempts
  * @param {number} [options.delayMs=1000] - Base delay between attempts in ms
  * @param {string} [options.description] - Label for logging
diff --git a/.github/scripts/api-helpers.js b/.github/scripts/api-helpers.js
index 6fa2a787a..885c2807c 100644
--- a/.github/scripts/api-helpers.js
+++ b/.github/scripts/api-helpers.js
@@ -7,6 +7,8 @@
  * This module addresses Issue R-1 from WorkflowSystemBugReport.md
  */
 
+const { withGithubApiRetry, calculateBackoffDelay } = require('./github_api_retry');
+
 const DEFAULT_MAX_RETRIES = 3;
 const DEFAULT_BASE_DELAY_MS = 1000;
 const DEFAULT_MAX_DELAY_MS = 30000;
@@ -58,23 +60,6 @@ function sleep(ms) {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
 
-/**
- * Calculate delay with exponential backoff and jitter
- * @param {number} attempt - Current attempt number (0-indexed)
- * @param {number} baseDelay - Base delay in milliseconds
- * @param {number} maxDelay - Maximum delay in milliseconds
- * @returns {number} Calculated delay with jitter
- */
-function calculateBackoffDelay(attempt, baseDelay = DEFAULT_BASE_DELAY_MS, maxDelay = DEFAULT_MAX_DELAY_MS) {
-  // Exponential backoff: baseDelay * 2^attempt
-  const exponentialDelay = baseDelay * Math.pow(2, attempt);
-  // Cap at max delay
-  const cappedDelay = Math.min(exponentialDelay, maxDelay);
-  // Add jitter (±25%)
-  const jitter = cappedDelay * 0.25 * (Math.random() * 2 - 1);
-  return Math.round(cappedDelay + jitter);
-}
-
 /**
  * Extract rate limit reset time from error or response headers
  * @param {Error|Object} errorOrResponse - Error object or API response
@@ -160,49 +145,25 @@ async function paginateWithBackoff(github, method, params, options = {}) {
     core = null,
   } = options;
 
-  let lastError = null;
-
-  for (let attempt = 0; attempt <= maxRetries; attempt++) {
-    try {
-      return await github.paginate(method, params);
-    } catch (error) {
-      lastError = error;
-
-      // Check if this is a rate limit error we should retry
-      const isRateLimit = isRateLimitError(error);
-      const isSecondaryLimit = isSecondaryRateLimitError(error);
-
-      if (!isRateLimit && !isSecondaryLimit) {
-        // Not a rate limit error, throw immediately
-        throw error;
-      }
-
-      // Check if we have retries left
-      if (attempt >= maxRetries) {
-        log(core, 'error', `Rate limit exhausted after ${maxRetries + 1} attempts. Giving up.`);
-        throw error;
-      }
-
-      // Calculate wait time
-      let delay;
-      if (isSecondaryLimit) {
-        // Secondary rate limits require longer waits
-        const resetTime = extractRateLimitReset(error);
-        delay = resetTime ? calculateWaitUntilReset(resetTime) : calculateBackoffDelay(attempt + 2, baseDelay, maxDelay);
-      } else {
-        const resetTime = extractRateLimitReset(error);
-        delay = resetTime ? calculateWaitUntilReset(resetTime) : calculateBackoffDelay(attempt, baseDelay, maxDelay);
-      }
-
-      const limitType = isSecondaryLimit ? 'secondary rate limit' : 'rate limit';
-      log(core, 'warning', `Hit ${limitType}. Retrying in ${Math.round(delay / 1000)}s (attempt ${attempt + 1}/${maxRetries + 1})`);
-
-      await sleep(delay);
+  // Use withGithubApiRetry for comprehensive transient error handling
+  return withGithubApiRetry(
+    () => github.paginate(method, params),
+    {
+      operation: 'read', // Pagination is typically a read operation
+      label: 'GitHub API pagination',
+      maxRetriesByOperation: {
+        read: maxRetries,
+        write: maxRetries,
+        dispatch: maxRetries,
+        admin: maxRetries,
+        unknown: maxRetries,
+      },
+      baseDelay,
+      maxDelay,
+      core,
+      backoffFn: calculateBackoffDelay,
     }
-  }
-
-  // Should not reach here, but just in case
-  throw lastError || new Error('Pagination failed with unknown error');
+  );
 }
 
 /**
@@ -225,43 +186,22 @@ async function withBackoff(apiCall, options = {}) {
     core = null,
   } = options;
 
-  let lastError = null;
-
-  for (let attempt = 0; attempt <= maxRetries; attempt++) {
-    try {
-      return await apiCall();
-    } catch (error) {
-      lastError = error;
-
-      const isRateLimit = isRateLimitError(error);
-      const isSecondaryLimit = isSecondaryRateLimitError(error);
-
-      if (!isRateLimit && !isSecondaryLimit) {
-        throw error;
-      }
-
-      if (attempt >= maxRetries) {
-        log(core, 'error', `Rate limit exhausted after ${maxRetries + 1} attempts. Giving up.`);
-        throw error;
-      }
-
-      let delay;
-      if (isSecondaryLimit) {
-        const resetTime = extractRateLimitReset(error);
-        delay = resetTime ? calculateWaitUntilReset(resetTime) : calculateBackoffDelay(attempt + 2, baseDelay, maxDelay);
-      } else {
-        const resetTime = extractRateLimitReset(error);
-        delay = resetTime ? calculateWaitUntilReset(resetTime) : calculateBackoffDelay(attempt, baseDelay, maxDelay);
-      }
-
-      const limitType = isSecondaryLimit ? 'secondary rate limit' : 'rate limit';
-      log(core, 'warning', `Hit ${limitType}. Retrying in ${Math.round(delay / 1000)}s (attempt ${attempt + 1}/${maxRetries + 1})`);
-
-      await sleep(delay);
-    }
-  }
-
-  throw lastError || new Error('API call failed with unknown error');
+  // Use withGithubApiRetry for comprehensive transient error handling
+  return withGithubApiRetry(apiCall, {
+    operation: 'read', // Default to read operation
+    label: 'GitHub API call',
+    maxRetriesByOperation: {
+      read: maxRetries,
+      write: maxRetries,
+      dispatch: maxRetries,
+      admin: maxRetries,
+      unknown: maxRetries,
+    },
+    baseDelay,
+    maxDelay,
+    core,
+    backoffFn: calculateBackoffDelay,
+  });
 }
 
 /**
@@ -375,7 +315,6 @@ module.exports = {
   isRateLimitError,
   isSecondaryRateLimitError,
   sleep,
-  calculateBackoffDelay,
   extractRateLimitReset,
   calculateWaitUntilReset,
 
diff --git a/.github/scripts/github_api_retry.js b/.github/scripts/github_api_retry.js
index a683acafa..0ca1c951c 100644
--- a/.github/scripts/github_api_retry.js
+++ b/.github/scripts/github_api_retry.js
@@ -1,7 +1,6 @@
 'use strict';
 
 const { classifyError, ERROR_CATEGORIES } = require('./error_classifier');
-const { calculateBackoffDelay } = require('./api-helpers');
 
 const DEFAULT_RETRY_LIMITS = Object.freeze({
   read: 3,
@@ -14,6 +13,23 @@ const DEFAULT_RETRY_LIMITS = Object.freeze({
 const DEFAULT_BASE_DELAY_MS = 1000;
 const DEFAULT_MAX_DELAY_MS = 30000;
 
+/**
+ * Calculate delay with exponential backoff and jitter
+ * @param {number} attempt - Current attempt number (0-indexed)
+ * @param {number} baseDelay - Base delay in milliseconds
+ * @param {number} maxDelay - Maximum delay in milliseconds
+ * @returns {number} Calculated delay with jitter
+ */
+function calculateBackoffDelay(attempt, baseDelay = DEFAULT_BASE_DELAY_MS, maxDelay = DEFAULT_MAX_DELAY_MS) {
+  // Exponential backoff: baseDelay * 2^attempt
+  const exponentialDelay = baseDelay * Math.pow(2, attempt);
+  // Cap at max delay
+  const cappedDelay = Math.min(exponentialDelay, maxDelay);
+  // Add jitter (±25%)
+  const jitter = cappedDelay * 0.25 * (Math.random() * 2 - 1);
+  return Math.round(cappedDelay + jitter);
+}
+
 function normaliseHeaders(headers) {
   if (!headers || typeof headers !== 'object') {
     return {};
@@ -135,4 +151,5 @@ module.exports = {
   resolveMaxRetries,
   computeRetryDelayMs,
   withGithubApiRetry,
+  calculateBackoffDelay,
 };
diff --git a/.github/scripts/keepalive_gate.js b/.github/scripts/keepalive_gate.js
index 66b62f307..b7d529503 100644
--- a/.github/scripts/keepalive_gate.js
+++ b/.github/scripts/keepalive_gate.js
@@ -1,6 +1,6 @@
 'use strict';
 
-const { paginateWithBackoff } = require('./api-helpers.js');
+const { paginateWithBackoff, withBackoff } = require('./api-helpers.js');
 
 const KEEPALIVE_LABEL = 'agents:keepalive';
 const AGENT_LABEL_PREFIX = 'agent:';
@@ -16,7 +16,7 @@ const ORCHESTRATOR_WORKFLOW_FILE = 'agents-70-orchestrator.yml';
 const WORKER_WORKFLOW_FILE = 'agents-72-codex-belt-worker.yml';
 const RECENT_COMPLETED_LOOKBACK_SECONDS = 300; // 5 minutes
 
-// Rate limit retry configuration
+// Rate limit retry configuration - now handled by api-helpers
 const RATE_LIMIT_MAX_RETRIES = 3;
 const RATE_LIMIT_BASE_DELAY_MS = 2000;
 
@@ -46,40 +46,6 @@ function isRateLimitError(error) {
   );
 }
 
-/**
- * Execute a GitHub API call with exponential backoff retry on rate limit errors.
- * @template T
- * @param {() => Promise<T>} fn - The API call to execute
- * @param {Object} [options]
- * @param {number} [options.maxRetries=3] - Maximum retry attempts
- * @param {number} [options.baseDelayMs=2000] - Base delay in milliseconds
- * @param {Object} [options.core] - GitHub Actions core for logging
- * @returns {Promise<T>}
- */
-async function withRateLimitRetry(fn, options = {}) {
-  const maxRetries = options.maxRetries ?? RATE_LIMIT_MAX_RETRIES;
-  const baseDelayMs = options.baseDelayMs ?? RATE_LIMIT_BASE_DELAY_MS;
-  const core = options.core;
-
-  let lastError;
-  for (let attempt = 0; attempt <= maxRetries; attempt++) {
-    try {
-      return await fn();
-    } catch (error) {
-      lastError = error;
-      if (!isRateLimitError(error) || attempt >= maxRetries) {
-        throw error;
-      }
-      const delay = baseDelayMs * Math.pow(2, attempt);
-      if (core?.info) {
-        core.info(`Rate limited, retrying in ${delay}ms (attempt ${attempt + 1}/${maxRetries})`);
-      }
-      await sleep(delay);
-    }
-  }
-  throw lastError;
-}
-
 function toInteger(value) {
   const parsed = Number.parseInt(value, 10);
   if (!Number.isFinite(parsed)) {
@@ -814,9 +780,9 @@ async function evaluateRunCapForPr({
 
   let pull;
   try {
-    const response = await withRateLimitRetry(
+    const response = await withBackoff(
       () => github.rest.pulls.get({ owner, repo, pull_number: number }),
-      { core }
+      { core, maxRetries: RATE_LIMIT_MAX_RETRIES, baseDelay: RATE_LIMIT_BASE_DELAY_MS }
     );
     pull = response.data;
   } catch (error) {
@@ -927,9 +893,9 @@ async function evaluateKeepaliveGate({ core, github, context, options = {} }) {
   let pr = pullRequest || null;
   if (!pr) {
     try {
-      const response = await withRateLimitRetry(
+      const response = await withBackoff(
         () => github.rest.pulls.get({ owner, repo, pull_number: prNumber }),
-        { core }
+        { core, maxRetries: RATE_LIMIT_MAX_RETRIES, baseDelay: RATE_LIMIT_BASE_DELAY_MS }
       );
       pr = response.data;
     } catch (error) {