fix: merge_agent_work blocks until complete instead of requiring polling

The mergemaster agent was burning all 30 turns polling get_merge_status
every 2 seconds while the merge pipeline takes ~2 minutes. It would
exhaust turns, exit, restart, and repeat — never seeing the result.

merge_agent_work now blocks with a 10-second internal poll loop and
returns the final result directly. The agent calls it once and gets
the answer. No more polling turns wasted.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

server/src/io/story_metadata.rs

@@ -133,10 +133,25 @@ fn build_metadata(front: FrontMatter) -> StoryMetadata {
 
 /// Write or update a `merge_failure:` field in the YAML front matter of a story file.
 ///
+/// The reason is stored as a quoted YAML string so that colons, hashes, and newlines
+/// in the failure message do not break front-matter parsing.
+/// If no front matter is present, this is a no-op (returns Ok).
+pub fn write_merge_failure(path: &Path, reason: &str) -> Result<(), String> {
+    let contents =
+        fs::read_to_string(path).map_err(|e| format!("Failed to read story file: {e}"))?;
+
+    // Produce a YAML-safe inline quoted string: collapse newlines, escape inner quotes.
+    let escaped = reason.replace('"', "\\\"").replace('\n', " ").replace('\r', "");
+    let yaml_value = format!("\"{escaped}\"");
+
+    let updated = set_front_matter_field(&contents, "merge_failure", &yaml_value);
+    fs::write(path, &updated).map_err(|e| format!("Failed to write story file: {e}"))?;
+    Ok(())
+}
+
 /// Write `review_hold: true` to the YAML front matter of a story file.
 ///
 /// Used to mark spikes that have passed QA and are waiting for human review.
-#[cfg(test)]
 pub fn write_review_hold(path: &Path) -> Result<(), String> {
     let contents =
         fs::read_to_string(path).map_err(|e| format!("Failed to read story file: {e}"))?;
@@ -149,7 +164,6 @@ pub fn write_review_hold(path: &Path) -> Result<(), String> {
 ///
 /// If front matter is present and contains the key, the line is removed.
 /// If no front matter or key is not found, the file is left unchanged.
-#[cfg(test)]
 pub fn clear_front_matter_field(path: &Path, key: &str) -> Result<(), String> {
     let contents =
         fs::read_to_string(path).map_err(|e| format!("Failed to read story file: {e}"))?;
@@ -227,12 +241,23 @@ pub fn set_front_matter_field(contents: &str, key: &str, value: &str) -> String
     result
 }
 
+/// Write `blocked: true` to the YAML front matter of a story file.
+///
+/// Used to mark stories that have exceeded the retry limit and should not
+/// be auto-assigned again.
+pub fn write_blocked(path: &Path) -> Result<(), String> {
+    let contents =
+        fs::read_to_string(path).map_err(|e| format!("Failed to read story file: {e}"))?;
+    let updated = set_front_matter_field(&contents, "blocked", "true");
+    fs::write(path, &updated).map_err(|e| format!("Failed to write story file: {e}"))?;
+    Ok(())
+}
+
 /// Write or update a `depends_on:` field in the YAML front matter of a story file.
 ///
 /// Serialises `deps` as an inline YAML sequence, e.g. `[477, 478]`.
 /// If `deps` is empty the field is removed.
 /// If no front matter is present, this is a no-op (returns Ok).
-#[cfg(test)]
 pub fn write_depends_on(path: &Path, deps: &[u32]) -> Result<(), String> {
     let contents =
         fs::read_to_string(path).map_err(|e| format!("Failed to read story file: {e}"))?;