diff --git a/architect/AGENTS.md b/architect/AGENTS.md
index 49d32b3..e705f23 100644
--- a/architect/AGENTS.md
+++ b/architect/AGENTS.md
@@ -10,9 +10,9 @@ converses with humans through PR comments.
 ## Role
 
 - **Input**: Vision issues from VISION.md, prerequisite tree from ops repo
-- **Output**: Sprint proposals as PRs on the ops repo, sub-issue files
+- **Output**: Sprint proposals as PRs on the ops repo (with embedded `## Sub-issues` blocks)
 - **Mechanism**: Bash-driven orchestration in `architect-run.sh`, pitching formula via `formulas/run-architect.toml`
-- **Identity**: `architect-bot` on Forgejo
+- **Identity**: `architect-bot` on Forgejo (READ-ONLY on project repo, write on ops repo only — #764)
 
 ## Responsibilities
 
@@ -24,16 +24,17 @@ converses with humans through PR comments.
    acceptance criteria and dependencies
 4. **Human conversation**: Respond to PR comments, refine sprint proposals based
    on human feedback
-5. **Sub-issue filing**: After design forks are resolved, file concrete sub-issues
-   for implementation
+5. **Sub-issue definition**: Define concrete sub-issues in the `## Sub-issues`
+   block of the sprint spec. Filing is handled by `filer-bot` after sprint PR
+   merge (#764)
 
 ## Formula
 
 The architect pitching is driven by `formulas/run-architect.toml`. This formula defines
 the steps for:
 - Research: analyzing vision items and prerequisite tree
-- Pitch: creating structured sprint PRs
-- Sub-issue filing: creating concrete implementation issues
+- Pitch: creating structured sprint PRs with embedded `## Sub-issues` blocks
+- Design Q&A: refining the sprint via PR comments after human ACCEPT
 
 ## Bash-driven orchestration
 
@@ -57,22 +58,31 @@ APPROVED review → start design questions (model posts Q1:, adds Design forks s
   ↓
 Answers received → continue Q&A (model processes answers, posts follow-ups)
   ↓
-All forks resolved → sub-issue filing (model files implementation issues)
+All forks resolved → finalize ## Sub-issues section in sprint spec
+  ↓
+Sprint PR merged → filer-bot files sub-issues on project repo (#764)
   ↓
 REJECT review → close PR + journal (model processes rejection, bash merges PR)
 ```
 
 ### Vision issue lifecycle
 
-Vision issues decompose into sprint sub-issues tracked via "Decomposed from #N" in sub-issue bodies. The architect automatically closes vision issues when all sub-issues are closed:
+Vision issues decompose into sprint sub-issues. Sub-issues are defined in the
+`## Sub-issues` block of the sprint spec (between `<!-- filer:begin -->` and
+`<!-- filer:end -->` markers) and filed by `filer-bot` after the sprint PR merges
+on the ops repo (#764).
 
-1. Before picking new vision issues, the architect checks each open vision issue
-2. For each, it queries merged sprint PRs — **only PRs whose title or body reference the specific vision issue** (matched via `#N` pattern, filtering out unrelated PRs that happen to close unrelated issues) (#735/#736)
-3. Extracts sub-issue numbers from those PRs, excluding the vision issue itself
-4. If all sub-issues are closed, posts a summary comment listing completed sub-issues (with an idempotency guard: checks both comment presence AND `.state == "closed"` — if the comment exists but the issue is still open, retries the close rather than returning early) (#737)
-5. The vision issue is then closed automatically
+Each filer-created sub-issue carries a `<!-- decomposed-from: #<vision>, sprint: <slug>, id: <id> -->`
+marker in its body for idempotency and traceability.
 
-This ensures vision issues transition from `open` → `closed` once their work is complete, without manual intervention. The #N-scoped matching prevents false positives where unrelated sub-issues would incorrectly trigger vision issue closure.
+The filer-bot (via `lib/sprint-filer.sh`) handles vision lifecycle:
+1. After filing sub-issues, adds `in-progress` label to the vision issue
+2. On each run, checks if all sub-issues for a vision are closed
+3. If all closed, posts a summary comment and closes the vision issue
+
+The architect no longer writes to the project repo — it is read-only (#764).
+All project-repo writes (issue filing, label management, vision closure) are
+handled by filer-bot with its narrowly-scoped `FORGE_FILER_TOKEN`.
 
 ### Session management
 
@@ -95,7 +105,9 @@ Run via `architect/architect-run.sh`, which:
   - Selects up to `pitch_budget` (3 - open architect PRs) remaining vision issues
   - For each selected issue, invokes stateless `claude -p` with issue body + context
   - Creates PRs directly from pitch content (no scratch files)
-- Agent is invoked only for response processing (ACCEPT/REJECT handling)
+- Agent is invoked for stateless pitch generation and response processing (ACCEPT/REJECT handling)
+- NOTE: architect-bot is read-only on the project repo (#764) — sub-issue filing
+  and in-progress label management are handled by filer-bot after sprint PR merge
 
 **Multi-sprint pitching**: The architect pitches up to 3 sprints per run. Bash handles all state management:
 - Fetches Forgejo API data (vision issues, open PRs, merged PRs)
@@ -120,4 +132,5 @@ empty file not created, just document it).
 - #100: Architect formula — research + design fork identification
 - #101: Architect formula — sprint PR creation with questions
 - #102: Architect formula — answer parsing + sub-issue filing
+- #764: Permission scoping — architect read-only on project repo, filer-bot files sub-issues
 - #491: Refactor — bash-driven design phase with stateful session resumption
diff --git a/lib/sprint-filer.sh b/lib/sprint-filer.sh
index e2b45a6..916d7c3 100755
--- a/lib/sprint-filer.sh
+++ b/lib/sprint-filer.sh
@@ -42,6 +42,31 @@ filer_log() {
 : "${FORGE_FILER_TOKEN:?sprint-filer.sh requires FORGE_FILER_TOKEN}"
 : "${FORGE_API:?sprint-filer.sh requires FORGE_API}"
 
+# ── Paginated Forgejo API fetch ──────────────────────────────────────────
+# Fetches all pages of a Forgejo API list endpoint and merges into one JSON array.
+# Args: api_path (e.g. /issues?state=all&type=issues)
+# Output: merged JSON array to stdout
+filer_api_all() {
+  local path_prefix="$1"
+  local sep page page_items count all_items="[]"
+  case "$path_prefix" in
+    *"?"*) sep="&" ;;
+    *) sep="?" ;;
+  esac
+  page=1
+  while true; do
+    page_items=$(curl -sf -H "Authorization: token ${FORGE_FILER_TOKEN}" \
+      "${FORGE_API}${path_prefix}${sep}limit=50&page=${page}" 2>/dev/null) || page_items="[]"
+    count=$(printf '%s' "$page_items" | jq 'length' 2>/dev/null) || count=0
+    [ -z "$count" ] && count=0
+    [ "$count" -eq 0 ] && break
+    all_items=$(printf '%s\n%s' "$all_items" "$page_items" | jq -s 'add')
+    [ "$count" -lt 50 ] && break
+    page=$((page + 1))
+  done
+  printf '%s' "$all_items"
+}
+
 # ── Parse sub-issues block from a sprint markdown file ───────────────────
 # Extracts the YAML-in-markdown between <!-- filer:begin --> and <!-- filer:end -->
 # Args: sprint_file_path
@@ -93,11 +118,36 @@ parse_subissues_block() {
 }
 
 # ── Extract vision issue number from sprint file ─────────────────────────
-# Looks for "## Vision issues" section with "#N" references
+# Looks for "#N" references specifically in the "## Vision issues" section
+# to avoid picking up cross-links or related-issue mentions earlier in the file.
+# Falls back to first #N in the file if no "## Vision issues" section found.
 # Args: sprint_file_path
 # Output: first vision issue number found
 extract_vision_issue() {
   local sprint_file="$1"
+
+  # Try to extract from "## Vision issues" section first
+  local in_section=false
+  local result=""
+  while IFS= read -r line; do
+    if [[ "$line" =~ ^##[[:space:]]+Vision[[:space:]]+issues ]]; then
+      in_section=true
+      continue
+    fi
+    # Stop at next heading
+    if [ "$in_section" = true ] && [[ "$line" =~ ^## ]]; then
+      break
+    fi
+    if [ "$in_section" = true ]; then
+      result=$(printf '%s' "$line" | grep -oE '#[0-9]+' | head -1 | tr -d '#')
+      if [ -n "$result" ]; then
+        printf '%s' "$result"
+        return 0
+      fi
+    fi
+  done < "$sprint_file"
+
+  # Fallback: first #N in the entire file
   grep -oE '#[0-9]+' "$sprint_file" | head -1 | tr -d '#'
 }
 
@@ -255,10 +305,9 @@ subissue_exists() {
 
   local marker="<!-- decomposed-from: #${vision_issue}, sprint: ${sprint_slug}, id: ${subissue_id} -->"
 
-  # Search for issues with this exact marker
+  # Search all issues (paginated) for the exact marker
   local issues_json
-  issues_json=$(curl -sf -H "Authorization: token ${FORGE_FILER_TOKEN}" \
-    "${FORGE_API}/issues?state=all&limit=50&type=issues" 2>/dev/null) || issues_json="[]"
+  issues_json=$(filer_api_all "/issues?state=all&type=issues")
 
   if printf '%s' "$issues_json" | jq -e --arg marker "$marker" \
     '[.[] | select(.body // "" | contains($marker))] | length > 0' >/dev/null 2>&1; then
@@ -444,8 +493,7 @@ check_and_close_completed_visions() {
   filer_log "Checking for vision issues with all sub-issues complete..."
 
   local vision_issues_json
-  vision_issues_json=$(curl -sf -H "Authorization: token ${FORGE_FILER_TOKEN}" \
-    "${FORGE_API}/issues?labels=vision&state=open&limit=100" 2>/dev/null) || vision_issues_json="[]"
+  vision_issues_json=$(filer_api_all "/issues?labels=vision&state=open")
 
   if [ "$vision_issues_json" = "[]" ] || [ "$vision_issues_json" = "null" ]; then
     filer_log "No open vision issues found"
@@ -453,8 +501,7 @@ check_and_close_completed_visions() {
   fi
 
   local all_issues
-  all_issues=$(curl -sf -H "Authorization: token ${FORGE_FILER_TOKEN}" \
-    "${FORGE_API}/issues?state=all&limit=200&type=issues" 2>/dev/null) || all_issues="[]"
+  all_issues=$(filer_api_all "/issues?state=all&type=issues")
 
   local vision_nums
   vision_nums=$(printf '%s' "$vision_issues_json" | jq -r '.[].number' 2>/dev/null) || return 0