diff --git a/docs/specification.md b/docs/specification.md
index 4accaad8b..82954102b 100644
--- a/docs/specification.md
+++ b/docs/specification.md
@@ -14912,14 +14912,14 @@ Including the raw model reasoning with `--show-reasoning`:
 ##### agents plan correct
 
 <div class="highlight"><pre><code><span style="color: cyan; font-weight: 600;">agents</span> plan correct <span style="color: cyan;">--mode</span> (<span style="color: magenta;">revert|append</span>) (<span style="color: magenta;"><span style="color: cyan;">--guidance</span>|<span style="color: yellow;">-g</span></span>) <span style="color: #66cc66;">&lt;GUIDANCE&gt;</span>
-                   [<span style="color: cyan;">--dry-run</span>] [<span style="color: cyan;">--yes</span>|<span style="color: yellow;">-y</span>] <span style="color: #66cc66;">&lt;DECISION_ID&gt;</span></code></pre></div>
+                   [<span style="color: cyan;">--dry-run</span>] [<span style="color: cyan;">--yes</span>|<span style="color: yellow;">-y</span>] <span style="color: #66cc66;">&lt;PLAN_ID|DECISION_ID&gt;</span></code></pre></div>
 
 **Purpose**
 Correct a decision either by reverting and re-executing or by appending a fix.
 
 **Arguments**
 
-- `<DECISION_ID>`: Decision ID.
+- `<PLAN_ID|DECISION_ID>`: The plan or decision to correct. When a plan ID is provided, the root decision of that plan is automatically selected as the correction target. When a decision ID is provided, that specific decision is targeted (backward-compatible behavior).
 - `--mode revert|append`: Correction mode.
 - `--guidance/-g TEXT`: Guidance text.
 - `--dry-run`: Show impact without executing.
@@ -15857,7 +15857,7 @@ Show error decisions with recovery hints and retry history for a plan.
     │   → Check sandbox permissions and retry execution                  │
     │     <span style="opacity: 0.7;">$ agents plan execute 01HXM8C2ZK4Q7C2B3F2R4VYV6J</span>              │
     │   → Revert to Strategize phase to adjust the plan                  │
-    │     <span style="opacity: 0.7;">$ agents plan correct --mode revert -g "..." &lt;DECISION_ID&gt;</span>     │
+     │     <span style="opacity: 0.7;">$ agents plan correct --mode revert -g "..." &lt;PLAN_ID|DECISION_ID&gt;</span>     │
     ╰────────────────────────────────────────────────────────────────────╯
 
     <span style="color: #66cc66; font-weight: 600;">✓ OK</span>
@@ -15882,7 +15882,7 @@ Show error decisions with recovery hints and retry history for a plan.
       → Check sandbox permissions and retry execution
         $ agents plan execute 01HXM8C2ZK4Q7C2B3F2R4VYV6J
       → Revert to Strategize phase to adjust the plan
-        $ agents plan correct --mode revert -g "..." <DECISION_ID>
+        $ agents plan correct --mode revert -g "..." <PLAN_ID|DECISION_ID>
 
     [OK]
     ```
@@ -15909,7 +15909,7 @@ Show error decisions with recovery hints and retry history for a plan.
         {
           "action": "revert",
           "message": "Revert to Strategize phase to adjust the plan",
-          "cli_command": "agents plan correct --mode revert -g \"...\" <DECISION_ID>"
+          "cli_command": "agents plan correct --mode revert -g \"...\" <PLAN_ID|DECISION_ID>"
         }
       ]
     }
@@ -18677,11 +18677,11 @@ Plan: 01KH29QDEE6DZTXKWNKCV8VP0F
 All corrections use the same unified command:
 
 <div class="highlight"><pre><code>
-agents plan correct &lt;decision_id&gt; <span style="color: cyan;">--mode</span>=&lt;mode&gt; <span style="color: cyan;">--guidance</span> <span style="color: #66cc66;">&quot;&lt;corrected decision text&gt;&quot;</span>
+agents plan correct &lt;plan_id_or_decision_id&gt; <span style="color: cyan;">--mode</span>=&lt;mode&gt; <span style="color: cyan;">--guidance</span> <span style="color: #66cc66;">&quot;&lt;corrected decision text&gt;&quot;</span>
 </code></pre></div>
 
 **Parameters:**
-* `<decision_id>`: The ULID of the decision to correct
+* `<plan_id_or_decision_id>`: Either a plan ULID (auto-selects the root decision) or a specific decision ULID. Using a plan ID is the most common usage — the root decision is automatically targeted. Providing a decision ID directly allows targeting any specific decision in the tree (backward-compatible).
 * `--mode`: Either `revert` (rollback and re-run) or `append` (add fix at end)
 * `--guidance`: Free-form text specifying what the correct decision should be