diff --git a/CHANGELOG.md b/CHANGELOG.md
index 542d3bb8b..444599d2c 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -6,6 +6,16 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Fixed — docs site strict build red since 2026-07-08 (root-relative links)
+
+`docs/guides/tuning.md` (PR #245) and `docs/PRODUCTION-CONTRACT.md` (PR #263)
+linked repo-root files (`BENCHMARK.md`, `RELEASES.md`,
+`.github/workflows/release.yml`) via relative paths that escape the MkDocs
+`docs/` tree — `mkdocs build --strict` aborts on the 5 resulting warnings, so
+the GitHub Pages deploy had failed on every main push since 2026-07-08.
+Converted the 5 links to absolute GitHub blob URLs; strict build verified
+clean locally. Docs-only.
+
 ### Changed — WAL v3 `wal_append` channel now preserves the caller's REAL record type end-to-end (K1a, storage-kernel M1 stage 1)
 
 `ShardDatabases::wal_append` / `try_wal_append_required` and
diff --git a/docs/PRODUCTION-CONTRACT.md b/docs/PRODUCTION-CONTRACT.md
index 0d346ab7e..f3bddcc6a 100644
--- a/docs/PRODUCTION-CONTRACT.md
+++ b/docs/PRODUCTION-CONTRACT.md
@@ -9,7 +9,7 @@ description: "Moon's v1.0 promises: SLOs, durability modes, supported platforms,
 what the `v1.0` release gate checks.
 **Last updated:** 2026-07-10
 **Milestone:** v0.6.1 Hygiene & reliability hotfix
-**Current release:** v0.6.0 (2026-07-08, PR #249) — see [`RELEASES.md`](../RELEASES.md)
+**Current release:** v0.6.0 (2026-07-08, PR #249) — see [`RELEASES.md`](https://github.com/pilotspace/moon/blob/main/RELEASES.md)
 **Roadmap:** [`docs/roadmap/ROADMAP.md`](roadmap/ROADMAP.md) §1 (proven strengths / reality check) is the
 authoritative source this ledger was verified against.
 
@@ -21,7 +21,7 @@ shipped). **No aspirational ticks** — a row is only ✅ if the linked evidence
 exist and do what the row claims, as of this refresh.
 
 `scripts/check-production-contract.sh` greps this file and is wired into
-[`.github/workflows/release.yml`](../.github/workflows/release.yml):
+[`.github/workflows/release.yml`](https://github.com/pilotspace/moon/blob/main/.github/workflows/release.yml):
 
 - On any `v0.x` tag it **reports** unticked GA-blocking rows but does not fail the build — Moon is
   pre-GA and the whole point of this ledger is to track the gap honestly.
@@ -46,7 +46,7 @@ in `.github/workflows/ci.yml`); production **benchmark numbers** still MUST come
 
 ## Performance
 
-Numbers are tracked in [`BENCHMARK.md`](../BENCHMARK.md) (KV pipelined/p=1, vector iso-recall vs
+Numbers are tracked in [`BENCHMARK.md`](https://github.com/pilotspace/moon/blob/main/BENCHMARK.md) (KV pipelined/p=1, vector iso-recall vs
 Qdrant/RediSearch, graph vs FalkorDB), refreshed per release rather than pinned as static SLOs in
 this document — no automated 24h HDR rig or Criterion CI regression gate exists yet (`PERF-01`/`PERF-02`
 below are unticked). Headline, evidence-cited numbers as of v0.6.0:
diff --git a/docs/guides/tuning.md b/docs/guides/tuning.md
index 0a8ffa0d3..4f297d06b 100644
--- a/docs/guides/tuning.md
+++ b/docs/guides/tuning.md
@@ -128,7 +128,7 @@ Moon's AOF write path is engineered so that durability is cheap at pipeline dept
 device-bound (not server-bound) when you demand fsync-per-write. The mechanics below are
 **automatic — there are no knobs to turn** — but knowing them tells you which policy fits
 your workload. Measured on GCE `c3-standard-8` (pd-ssd), `--shards 2`, vs Redis 7.0.15;
-full matrix in [`BENCHMARK.md` §7.3](../../BENCHMARK.md).
+full matrix in [`BENCHMARK.md` §7.3](https://github.com/pilotspace/moon/blob/main/BENCHMARK.md).
 
 - **`everysec` (default): a win, not just free.** At pipeline depth SET runs **~1.32×
   Redis** (the writer coalesces each batch into one `write_all` and polls its queue
@@ -164,7 +164,7 @@ faster than a subscriber's socket drains, the whole burst is delivered in one `w
 instead of one syscall per message. In practice this means a fast publisher fanning out to
 many subscribers is no longer syscall-bound: measured fan-out delivery reached **5.09M
 msg/s (≈1.04× Redis) with zero drops**, versus near-total message loss on a per-message
-write path (see [`BENCHMARK.md` §7.3](../../BENCHMARK.md)). There are no knobs to turn.
+write path (see [`BENCHMARK.md` §7.3](https://github.com/pilotspace/moon/blob/main/BENCHMARK.md)). There are no knobs to turn.
 
 The one thing to know: each subscriber has a bounded in-flight queue (256 messages). A
 subscriber that stays slower than the publish rate will still have messages **dropped** —