Add AI ethics review checklist as PR template

Adds .github/PULL_REQUEST_TEMPLATE.md with a mandatory AI safety and
security checklist covering Input Validation, Least Privilege, Credential
Safety, Human Oversight, AI Output Safety, and Testing — codifying the
principles from AI_PRINCIPLES.md into every PR review. Closes #13.

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

Author: 2 people

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,29 @@
+## Changes
+<!-- Describe what this PR does -->
+
+## AI Safety & Security Checklist
+<!-- Check all that apply. If a box is unchecked, explain why in the PR description. -->
+
+### Input Validation
+- [ ] All external inputs (user messages, API responses, file contents) are validated
+- [ ] Path operations stay within workspace boundaries
+
+### Least Privilege
+- [ ] New tools/features request only the permissions they need
+- [ ] No unnecessary filesystem, network, or system access added
+
+### Credential Safety
+- [ ] No API keys, tokens, or passwords in code, logs, or LLM context
+- [ ] Sensitive data masked in all log output
+
+### Human Oversight
+- [ ] Destructive or irreversible actions require user confirmation
+- [ ] Agent announces intent before high-risk operations
+
+### AI Output Safety
+- [ ] LLM outputs are validated before acting on them
+- [ ] Tool outputs are sanitized before re-entering LLM context
+
+### Testing
+- [ ] Security tests added/updated for changes
+- [ ] No test coverage decrease