diff --git a/.claude/commands/article-audit.md b/.claude/commands/article-audit.md
index 6fc07485..d46e3507 100644
--- a/.claude/commands/article-audit.md
+++ b/.claude/commands/article-audit.md
@@ -198,9 +198,12 @@ One of:
 - **Open an issue for a genuine gap** — same shape, but the work is larger.
   Flag if it's an epic.
 
-If the answer is "open an issue," **ask the user before running
-`gh issue create`.** They might want to tweak wording, add labels, or defer.
-Don't just create issues on their behalf.
+If the answer is "open an issue," **first search for an existing one** —
+`gh issue list --state open --search "<keywords>"` — and if a live issue
+already covers the topic, extend/comment on it instead of opening a
+near-duplicate (see CLAUDE.md "Search before you open an issue"). Then **ask
+the user before running `gh issue create`.** They might want to tweak wording,
+add labels, or defer. Don't just create issues on their behalf.
 
 ## Style notes
 
diff --git a/.claude/commands/feature-spec.md b/.claude/commands/feature-spec.md
index 737ab451..2793cb5e 100644
--- a/.claude/commands/feature-spec.md
+++ b/.claude/commands/feature-spec.md
@@ -188,6 +188,10 @@ Produce:
    Use the project's [work-item Issue Form fields](../../.github/ISSUE_TEMPLATE/work-item.yml)
    if possible (What / Why / Acceptance / Notes). Suggest labels:
    `type/feature`, the relevant `area/*`, `priority/*` if known.
+   **Before creating: search for an existing issue** —
+   `gh issue list --state open --search "<keywords>"` — and extend a live one
+   rather than opening a near-duplicate, especially for epics (see CLAUDE.md
+   "Search before you open an issue").
 
 2. **Optional ADR draft** — if the significance check returned yes.
 
diff --git a/CLAUDE.md b/CLAUDE.md
index 1b3ebb83..8461780f 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -218,6 +218,8 @@ When you find an antipattern, rule, or specification worth encoding, write to th
 
 **Deferral surface (NOT part of the encoding loop).** GitHub Issues (`rule-encoding-deferred` label) tracks findings where code shipped but the encoding hasn't yet. **The issue is a placeholder — a TODO that ensures the encoding eventually happens.** The issue itself is NOT the encoding; the PR that adds the rule to one of the five surfaces above is. Closed issues are not read again in future sessions; the rules live in the surfaces, not in the issue tracker.
 
+**Search before you open an issue (especially epics).** Before `gh issue create`, run `gh issue list --state open --search "<keywords>"` (and grep titles for the topic) — if an open issue already covers it, comment on or extend that one instead of opening a near-duplicate. Duplicate epics are the costly case: two "frontend" epics fragment progress tracking and the close-as-we-go discipline silently breaks. If a genuine overlap is found after the fact, consolidate to ONE canonical issue (fold the better framing in, close the rest as `duplicate` with a pointer) rather than leaving parallel trackers. Precedent: #129/#130/#102 frontend-epic consolidation, 2026-06-12.
+
 The threshold for encoding is the same as the Debugging Discipline rule: **if the next person could repeat the mistake (or re-derive the rule from first principles), the rule belongs in writing.** Don't encode trivial style nits or one-off mistakes; do encode security patterns, performance traps, concurrency hazards, distributed-systems gotchas, anti-IDOR patterns, anti-IDOR-test patterns, outbox traps, anything cross-cutting.
 
 When you push a fix PR for a real finding, the *fix itself* lives in the PR but the *rule* lives in `.claude/`. The two should land together when feasible (single PR with both), or as paired PRs when separation is cleaner. **A merged fix PR without the corresponding rule is a half-finished job** — the next instance of the same antipattern will slip through.