WIP (experiment DO NOT MERGE) Add WLED secure coding AI review instruction guides

docs/secure-short.instructions.md

@@ -0,0 +1,44 @@
+---
+applyTo: "**/*.{cpp,h,hpp,ino,js,htm,html,css,yml,yaml}"
+description: "WLED strict-mode security review: low-noise 24-rule checklist."
+---
+
+# WLED Security Review — Strict Mode (Low Noise)
+
+Use these 24 rules for automated reviews with minimal false positives.
+
+## CRITICAL Rules
+
+1. **No unchecked buffer copies** (`memcpy`, `strcpy`, `sprintf`) in firmware paths.
+2. **No user-controlled format strings** in `DEBUG_PRINTF*` and similar logging APIs.
+3. **Validate all external input** (HTTP/JSON/UDP/serial) before index/length/pin usage.
+4. **Auth required for every state-changing endpoint**.
+5. **No fail-open on parse/allocation errors** for config/state updates.
+6. **No DOM XSS sinks with untrusted data** (`innerHTML`, unsafe HTML insertion).
+7. **No dynamic code execution** (`eval`, `new Function`, string timers).
+8. **No hardcoded secrets/credentials/tokens/keys** in committed files.
+9. **No sensitive data in logs** (passwords, tokens, Wi-Fi secrets, auth headers).
+10. **No secret exposure in workflows/log output**.
+11. **No unsafe third-party GitHub Action pinning** (`@main`/`@master` disallowed).
+12. **No untrusted expression interpolation in workflow shell commands**.
+
+## IMPORTANT Rules
+
+13. Check integer overflow risks in size/index arithmetic.
+14. Reject repeated heap allocation churn in hot render/effect loops.
+15. Avoid repeated `String` growth in hot paths; prefer bounded/pre-allocated buffers.
+16. Ensure UI validation is mirrored by firmware-side validation.
+17. Require strict origin checks for `postMessage` listeners.
+18. Disallow untrusted redirect/navigation targets.
+19. Prevent verbose error responses that leak internals.
+20. Review new dependencies for typosquatting and known vulnerability risk.
+21. Keep workflow `permissions` least-privilege.
+22. Verify new `WLED_ENABLE_*` / `WLED_DISABLE_*` names are valid known flags.
+23. Ensure new privileged behavior is not enabled by insecure defaults.
+24. Preserve safe behavior under malformed inputs and low-memory conditions.
+
+## Reviewer Output Format
+
+- Report only findings mapped to rules 1–24.
+- Include severity, exact file and line, and one concrete fix direction.
+- Prioritize CRITICAL findings before IMPORTANT findings.

docs/securecode.instructions.md

@@ -0,0 +1,147 @@
+---
+applyTo: "**/*.{cpp,h,hpp,ino,js,htm,html,css,yml,yaml}"
+description: "WLED-focused security review guide based on OWASP Top 10 for embedded firmware and web UI."
+---
+
+# WLED Security Review Standards (Embedded + Web UI)
+
+Use this guide for AI-assisted code reviews in:
+- `/home/runner/work/WLED/WLED/wled00/`
+- `/home/runner/work/WLED/WLED/usermods/`
+- `/home/runner/work/WLED/WLED/.github/workflows/`
+
+Ignore sections wrapped in `<!-- HUMAN_ONLY_START --> ... <!-- HUMAN_ONLY_END -->` in repo docs when applying review criteria.
+
+## Severity
+
+- **CRITICAL** — exploitable vulnerability; block merge.
+- **IMPORTANT** — meaningful risk; fix before or with merge when practical.
+- **SUGGESTION** — defense-in-depth; track for follow-up.
+
+## Scope (WLED-relevant)
+
+Prioritize:
+- C++ memory safety and input validation
+- Auth and access checks for state-changing HTTP/JSON APIs
+- XSS and DOM safety in `wled00/data/*`
+- Secrets handling and secure logging
+- Dependency and GitHub Actions supply-chain hygiene
+- Fail-safe behavior on constrained devices
+
+De-prioritize unless explicitly introduced by a PR:
+- SQL/NoSQL checks, JWT/OAuth flows, GraphQL-specific checks, generic backend framework checks not used by WLED.
+
+## Firmware Security (C++, OWASP A01/A04/A05/A10)
+
+### FW1: Unsafe buffer operations
+- **Severity**: CRITICAL
+- Flag `strcpy`, `sprintf`, unchecked `memcpy`, unchecked pointer arithmetic.
+- Require explicit bounds checks and length validation.
+
+### FW2: Format-string injection
+- **Severity**: CRITICAL
+- Do not pass untrusted input as a format string to `DEBUG_PRINTF*` or similar APIs.
+
+### FW3: Integer overflow in length and offset math
+- **Severity**: IMPORTANT
+- Review `count * size`, index math, narrowing casts before allocations or copies.
+
+### FW4: Unvalidated external input
+- **Severity**: CRITICAL
+- Validate and clamp external values from HTTP/JSON/UDP/serial before use as lengths, indices, IDs, or pin references.
+
+### FW5: Missing auth checks on state-changing endpoints
+- **Severity**: CRITICAL
+- Any endpoint that changes device state/config must enforce configured auth policy.
+
+### FW6: Fail-open behavior after parse or allocation errors
+- **Severity**: IMPORTANT
+- On error, reject update and preserve safe previous state.
+
+### FW7: Heap churn in hot paths
+- **Severity**: IMPORTANT
+- Avoid repeated dynamic allocation in render/effect loops; prefer pre-allocation and reuse.
+
+### FW8: Unsafe use of `String` in performance-critical paths
+- **Severity**: IMPORTANT
+- In hot paths, avoid repeated `String` growth; reserve or use fixed buffers.
+
+### FW9: Unsafe feature flag names
+- **Severity**: IMPORTANT
+- Verify all new `WLED_ENABLE_*`/`WLED_DISABLE_*` names are valid known flags; typos silently alter build behavior.
+
+## Web UI Security (`wled00/data/*`, OWASP A01/A02/A05)
+
+### WEB1: DOM XSS through `innerHTML`
+- **Severity**: CRITICAL
+- Prefer `textContent`; if HTML is required, sanitize trusted content path explicitly.
+
+### WEB2: Dynamic code execution
+- **Severity**: CRITICAL
+- Reject `eval`, `new Function`, and string-based timer execution.
+
+### WEB3: `postMessage` without origin validation
+- **Severity**: IMPORTANT
+- Require strict origin allowlist checks before processing payloads.
+
+### WEB4: Unsafe redirects/navigation
+- **Severity**: IMPORTANT
+- Do not navigate directly from untrusted query/input without relative-path or allowlist checks.
+
+### WEB5: Client-only validation
+- **Severity**: IMPORTANT
+- UI validation is not sufficient; equivalent firmware-side validation is required.
+
+### WEB6: Direct DOM insertion from fetched/config data
+- **Severity**: IMPORTANT
+- Treat fetched and config-derived strings as untrusted unless proven otherwise.
+
+## Secrets and Logging (OWASP A04/A09/A10)
+
+### SEC1: Hardcoded secrets and credentials
+- **Severity**: CRITICAL
+- Reject committed API keys, passwords, tokens, private keys, or test backdoors.
+
+### SEC2: Sensitive values in logs
+- **Severity**: CRITICAL
+- Do not log passwords, tokens, Wi-Fi keys, auth headers, or full sensitive payloads.
+
+### SEC3: Insecure defaults
+- **Severity**: IMPORTANT
+- Reject new default credentials or insecure auto-enable behavior for privileged functions.
+
+### SEC4: Overly detailed error responses
+- **Severity**: IMPORTANT
+- Avoid exposing stack traces or internal details to API/UI consumers.
+
+## Supply Chain and CI/CD (OWASP A03/A08)
+
+### SC1: New dependency risk
+- **Severity**: IMPORTANT
+- Review new npm/pip/PlatformIO dependencies for legitimacy, pinning, and known vulnerabilities.
+
+### SC2: Workflow permission hardening regressions
+- **Severity**: IMPORTANT
+- Check for broad `permissions`, unpinned third-party actions, or unsafe secret exposure.
+
+### SC3: Script injection in workflows
+- **Severity**: IMPORTANT
+- Avoid direct interpolation of untrusted `${{ github.event.* }}` values in `run` commands.
+
+## Reviewer Checklist
+
+- [ ] No new memory-safety hazards (bounds, overflow, unsafe copies/format strings)
+- [ ] External input is validated and range-clamped before use
+- [ ] State-changing API paths enforce auth policy
+- [ ] Web UI changes avoid unsafe DOM execution/injection patterns
+- [ ] No secrets added; no sensitive logging introduced
+- [ ] Error handling remains fail-safe and non-leaky
+- [ ] Dependency/workflow changes are supply-chain safe
+- [ ] Feature-flag names are valid and not typoed
+
+## AI Review Behavior
+
+- Prefer concrete, file/line-specific findings over generic guidance.
+- Prioritize **CRITICAL** and **IMPORTANT** findings.
+- Skip irrelevant framework checks not used by WLED.
+- If control-flow trust is unclear, ask for clarification instead of guessing.