refactor: extract review logic into standalone script

Move all review logic from inline YAML to .github/scripts/perspective-review.sh.
Workflow YAML is now minimal — just triggers, permissions, and script invocation.
This fixes YAML parsing issues with complex heredocs and inline bash.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: anakrish

.github/scripts/perspective-review.sh

@@ -0,0 +1,207 @@
+#!/usr/bin/env bash
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+#
+# Multi-perspective PR review using GitHub Models API.
+# Called by the perspective-review.yml workflow.
+#
+# Usage: perspective-review.sh <repo> <pr_number>
+# Requires: GITHUB_TOKEN env var, jq, gh CLI
+
+set -euo pipefail
+
+REPO="$1"
+PR_NUMBER="$2"
+
+echo "=== Perspective Review for ${REPO}#${PR_NUMBER} ==="
+
+# Step 1: Get changed files
+echo "Fetching changed files..."
+gh api "repos/${REPO}/pulls/${PR_NUMBER}/files" \
+  --jq '.[].filename' > /tmp/changed_files.txt
+echo "Changed files: $(wc -l < /tmp/changed_files.txt)"
+
+# Step 2: Get the diff (truncate to ~60KB for token limits)
+echo "Fetching diff..."
+gh api "repos/${REPO}/pulls/${PR_NUMBER}" \
+  -H "Accept: application/vnd.github.v3.diff" \
+  | head -c 60000 > /tmp/pr_diff.txt
+echo "Diff size: $(wc -c < /tmp/pr_diff.txt) bytes"
+
+# Step 3: Select perspectives based on changed paths
+PERSPECTIVES="reliability-engineer,test-engineer"
+
+if grep -qE 'src/builtins/' /tmp/changed_files.txt 2>/dev/null; then
+  PERSPECTIVES="${PERSPECTIVES},semantics-expert,red-teamer"
+fi
+if grep -qE 'src/(value|number)' /tmp/changed_files.txt 2>/dev/null; then
+  PERSPECTIVES="${PERSPECTIVES},semantics-expert"
+fi
+if grep -qE 'bindings/|src/.*ffi' /tmp/changed_files.txt 2>/dev/null; then
+  PERSPECTIVES="${PERSPECTIVES},architect,api-steward"
+fi
+if grep -qE 'Cargo\.(toml|lock)' /tmp/changed_files.txt 2>/dev/null; then
+  PERSPECTIVES="${PERSPECTIVES},security-auditor,architect"
+fi
+if grep -qE 'src/(interpreter|rvm|compiler|scheduler)' /tmp/changed_files.txt 2>/dev/null; then
+  PERSPECTIVES="${PERSPECTIVES},semantics-expert,performance-engineer"
+fi
+
+# Deduplicate
+PERSPECTIVES=$(echo "$PERSPECTIVES" | tr ',' '\n' | sort -u | tr '\n' ',' | sed 's/,$//')
+echo "Selected perspectives: ${PERSPECTIVES}"
+
+# Step 4: Build context
+echo "Building context..."
+KNOWLEDGE_CTX=""
+for f in builtin-system value-semantics policy-evaluation-security ffi-boundary; do
+  kf="docs/knowledge/${f}.md"
+  if [ -f "$kf" ]; then
+    KNOWLEDGE_CTX="${KNOWLEDGE_CTX}
+--- ${f}.md ---
+$(head -c 4000 "$kf")
+"
+  fi
+done
+
+DIFF_CONTENT=$(cat /tmp/pr_diff.txt)
+
+# Step 5: Run each perspective
+ALL_FINDINGS=""
+
+for perspective in $(echo "$PERSPECTIVES" | tr ',' ' '); do
+  echo "--- Reviewing: ${perspective} ---"
+
+  # Load agent instructions
+  AGENT_INSTRUCTIONS=""
+  agent_file=".github/agents/${perspective}.agent.md"
+  if [ -f "$agent_file" ]; then
+    AGENT_INSTRUCTIONS=$(head -c 4000 "$agent_file")
+  fi
+
+  # Format display name
+  DISPLAY_NAME=$(echo "$perspective" | sed 's/-/ /g' | awk '{for(i=1;i<=NF;i++) $i=toupper(substr($i,1,1)) substr($i,2)}1')
+
+  # Build prompt as a temp file to avoid heredoc/quoting issues
+  cat > /tmp/review_prompt.txt <<EOF
+You are reviewing a pull request from the perspective of a ${DISPLAY_NAME}.
+
+Your agent instructions:
+${AGENT_INSTRUCTIONS}
+
+Relevant knowledge context:
+${KNOWLEDGE_CTX}
+
+PR Diff:
+${DIFF_CONTENT}
+
+Respond with a JSON array of findings. Each finding must have these fields:
+- "severity": one of "critical", "important", "suggestion"
+- "summary": a single sentence suitable as a GitHub issue title
+- "file": the file path (from the diff)
+- "line": approximate line number (from the diff), or null
+- "explanation": 2-4 sentences explaining the issue
+
+If you find no issues from this perspective, return an empty array: []
+
+Return ONLY valid JSON — no markdown fences, no extra text.
+EOF
+
+  PROMPT_CONTENT=$(cat /tmp/review_prompt.txt)
+
+  # Call GitHub Models API using jq for safe JSON encoding
+  RESPONSE=$(jq -n \
+    --arg model "openai/gpt-4o-mini" \
+    --arg prompt "$PROMPT_CONTENT" \
+    '{
+      model: $model,
+      messages: [
+        {role: "system", content: "You are a code reviewer. Return findings as a JSON array only."},
+        {role: "user", content: $prompt}
+      ],
+      temperature: 0.1
+    }' | curl -s -X POST "https://models.github.ai/inference/chat/completions" \
+      -H "Authorization: Bearer ${GITHUB_TOKEN}" \
+      -H "Content-Type: application/json" \
+      -d @- 2>/dev/null || echo '{"error": "API call failed"}')
+
+  # Extract content
+  CONTENT=$(echo "$RESPONSE" | jq -r '.choices[0].message.content // empty' 2>/dev/null || true)
+
+  if [ -z "$CONTENT" ]; then
+    ERROR_MSG=$(echo "$RESPONSE" | jq -r '.error // .message // "Unknown error"' 2>/dev/null || echo "Unknown error")
+    echo "  API error: ${ERROR_MSG}"
+    continue
+  fi
+
+  # Try to parse as JSON array and render as tagged markdown
+  RENDERED=$(echo "$CONTENT" | jq -r --arg perspective "$DISPLAY_NAME" '
+    if type == "array" then
+      .[] |
+      "**[\($perspective)]** " +
+      (if .severity == "critical" then "🔴 critical" elif .severity == "important" then "🟠 important" else "🔵 suggestion" end) +
+      "\n> " + .summary + "\n\n" + .explanation +
+      (if .file then "\n\n📁 `" + .file + "`" + (if .line then ":" + (.line | tostring) else "" end) else "" end) +
+      "\n\n---\n"
+    else
+      empty
+    end
+  ' 2>/dev/null || true)
+
+  if [ -n "$RENDERED" ]; then
+    FINDING_COUNT=$(echo "$CONTENT" | jq 'if type == "array" then length else 0 end' 2>/dev/null || echo 0)
+    echo "  Found ${FINDING_COUNT} findings"
+    ALL_FINDINGS="${ALL_FINDINGS}${RENDERED}"
+  else
+    echo "  No findings (or unparseable response)"
+  fi
+done
+
+# Step 6: Build the final comment
+MARKER="<!-- perspective-review-bot -->"
+
+if [ -n "$ALL_FINDINGS" ]; then
+  cat > /tmp/review_comment.md <<EOF
+${MARKER}
+## 🔍 Perspective Review
+
+Automated multi-perspective review of this PR.
+Each finding is tagged with the perspective that identified it.
+
+---
+
+$(echo -e "$ALL_FINDINGS")
+
+<sub>Generated by perspective-review workflow • Perspectives: ${PERSPECTIVES}</sub>
+EOF
+else
+  cat > /tmp/review_comment.md <<EOF
+${MARKER}
+## 🔍 Perspective Review
+
+✅ No significant findings from the selected perspectives.
+
+<sub>Generated by perspective-review workflow • Perspectives: ${PERSPECTIVES}</sub>
+EOF
+fi
+
+# Step 7: Upsert the comment (update existing or create new)
+echo "Posting review comment..."
+EXISTING_ID=$(gh api "repos/${REPO}/issues/${PR_NUMBER}/comments" \
+  --jq ".[] | select(.body | contains(\"${MARKER}\")) | .id" \
+  | head -1 || true)
+
+COMMENT_BODY=$(cat /tmp/review_comment.md)
+
+if [ -n "$EXISTING_ID" ]; then
+  gh api "repos/${REPO}/issues/comments/${EXISTING_ID}" \
+    -X PATCH \
+    -f body="$COMMENT_BODY"
+  echo "Updated existing comment ${EXISTING_ID}"
+else
+  gh api "repos/${REPO}/issues/${PR_NUMBER}/comments" \
+    -f body="$COMMENT_BODY"
+  echo "Created new review comment"
+fi
+
+echo "=== Review complete ==="