diff --git a/.codespellrc b/.codespellrc
index c554845f3..7f9b6d589 100644
--- a/.codespellrc
+++ b/.codespellrc
@@ -56,7 +56,9 @@
 
 # ans - bash and powershell variable short for answer
 
-ignore-words-list = numer,wit,aks,edn,ser,ois,gir,rouge,categor,aline,ative,afterall,deques,dateA,dateB,TE,FillIn,alle,vai,LOD,InOut,pixelX,aNULL,Wee,Sherif,queston,Vertexes,nin,FO,CAF,Parth,ans
+# GUD - "Guideline" identifier prefix in the create-implementation-plan skill spec (alongside REQ, SEC, CON, PAT, etc.)
+
+ignore-words-list = numer,wit,aks,edn,ser,ois,gir,rouge,categor,aline,ative,afterall,deques,dateA,dateB,TE,FillIn,alle,vai,LOD,InOut,pixelX,aNULL,Wee,Sherif,queston,Vertexes,nin,FO,CAF,Parth,ans,gud
 
 # Skip certain files and directories
 
diff --git a/skills/create-implementation-plan/SKILL.md b/skills/create-implementation-plan/SKILL.md
index 08a91438b..4540a479c 100644
--- a/skills/create-implementation-plan/SKILL.md
+++ b/skills/create-implementation-plan/SKILL.md
@@ -60,6 +60,34 @@ All implementation plans must strictly adhere to the following template. Each se
 - All identifier prefixes must follow the specified format
 - Tables must include all required columns
 - No placeholder text may remain in the final output
+- **Identifiers must be uniquely declared.** Every identifier (`REQ-NNN`, `SEC-NNN`, `CON-NNN`, `GUD-NNN`, `PAT-NNN`, `GOAL-NNN`, `TASK-NNN`, `ALT-NNN`, `DEP-NNN`, `FILE-NNN`, `TEST-NNN`, `RISK-NNN`, `ASSUMPTION-NNN`) must be **declared exactly once**. A declaration is where the identifier introduces a row: the leading cell in a TASK/GOAL table row, or the bolded prefix in a bullet line like `- **REQ-001**: ...`. The same identifier may then appear any number of times as a **reference** elsewhere in the plan (a `TASK` body citing a `REQ`, one `TASK` citing another `TASK`, the Dependencies section pointing at a `DEP` already declared upstream, etc.). References are expected and not collisions.
+
+## Identifier Uniqueness Check
+
+Run these checks before finalizing the plan. Checks (1) and (2) target declarations and must return zero rows. Check (3) is a broad informational scan: it will surface valid references too, so use it for awareness rather than as a gate.
+
+```bash
+# Set PLAN_FILE to the plan being validated.
+PLAN_FILE="/plan/<purpose>-<component>-<version>.md"
+
+# 1) Duplicate TASK / GOAL declarations in table rows.
+grep -oE '\| (TASK|GOAL)-[0-9]+ \|' "$PLAN_FILE" \
+  | sed -E 's/.*((TASK|GOAL)-[0-9]+).*/\1/' \
+  | sort | uniq -d
+
+# 2) Duplicate declaration IDs in bullet-style spec lines.
+grep -oE '^- \*\*(REQ|SEC|CON|GUD|RISK|ASSUMPTION|TASK|GOAL|FILE|TEST|PAT|ALT|DEP)-[0-9]+\*\*:' "$PLAN_FILE" \
+  | sed -E 's/^- \*\*([A-Z]+-[0-9]+)\*\*:.*/\1/' \
+  | sort | uniq -d
+
+# 3) Broad duplicate scan (diagnostic only; may include valid references).
+grep -oE '(REQ|SEC|CON|GUD|RISK|ASSUMPTION|TASK|GOAL|FILE|TEST|PAT|ALT|DEP)-[0-9]+' "$PLAN_FILE" \
+  | sort | uniq -d
+```
+
+Prerequisites: a POSIX-compatible shell (`sh` / `bash`) with `grep`, `sed`, `sort`, and `uniq`. On Windows without these tools, use equivalent platform-native commands and preserve the same declaration-vs-reference logic.
+
+If check (1) or (2) returns any row, re-number the duplicate so each identifier is declared exactly once, then re-run the checks until both are empty.
 
 ## Status