| name | security-checklist |
| description | The security review checklist (risk / validation / abuse / data) with this repo's anchors. Load when reviewing security-sensitive changes — capabilities, IPC, deps, secrets, AI, privacy, rate-limits. |
Security checklist
Authoritative: docs/knowledge/security-rules.md. Severity bias for security/data findings = round UP.
Risk assessment (every change)
- What assets / user data are affected? What attack surface changes? What abuse opportunities open?
Validation
- Inputs validated · outputs sanitized · permissions minimized · secrets protected · errors handled securely · logging reviewed (no secrets/PII in logs) · dependencies reviewed.
Abuse / cost (DoS & spend)
- Rate limits / request throttling present? AI usage + cost caps? Export/resource limits? Can a user spam, exhaust CPU/memory, or run up API spend?
AI security
- Can user input manipulate system prompts (injection)? Can sensitive data leak into prompts? Can AI reach unintended tools? Is AI output validated?
Data
- Resume/PII protected · temp files cleaned up · export files secured · local storage/cache secured · retention & deletion honored (GDPR).
Desktop / supply chain
tauri.conf.json CSP intact (incl. Ollama 127.0.0.1:11434) · capabilities/default.json least-privilege · updater signing key + latest.json integrity · deny.toml / cargo audit / pnpm audit clean · new deps license-checked.
External standards & best-practices (verified 2026-06-19)
Re-baseline to current editions: OWASP Top 10 2025 (was 2021), ASVS 5.0 (was 4.x), SLSA v1.2 (was v1.0), LLM Top 10 2025, CWE Top 25 2025. Cite these, not older ones.
- OWASP Top 10:2025 (RC Nov 2025 — confirm final before locking ranks) https://owasp.org/Top10/2025/ — new vs 2021:
A03 Software Supply-Chain Failures (expands old A06) and A10 Mishandling of Exceptional Conditions; SSRF folded into A01; A02 Security Misconfiguration rose to #2.
- OWASP ASVS 5.0.0 (2025-05-30) — verify against L1/L2/L3. https://github.com/OWASP/ASVS
- OWASP LLM Top 10 (2025) — LLM01 Prompt Injection · LLM02 Sensitive-Info · LLM05 Improper Output Handling · LLM06 Excessive Agency · LLM07 System-Prompt Leakage. https://genai.owasp.org/llm-top-10/
- Supply chain (now OWASP A03): target SLSA Build L3 (https://slsa.dev/spec/v1.2/); commit
pnpm-lock.yaml + CI --frozen-lockfile; cargo audit/cargo deny vs RUSTSEC daily. ⚠️ Provenance ≠ safety — the 2025–26 Shai-Hulud npm worm shipped malware with valid SLSA L3 provenance via hijacked OIDC tokens → pin exact versions, --ignore-scripts, scope-lock publish tokens. https://unit42.paloaltonetworks.com/npm-supply-chain-attack/
- Frameworks: NIST SSDF (SP 800-218) + CISA Secure-by-Design as the baseline.
- Secrets → OS keychain (
tauri-plugin-stronghold/keyring), never config/logs. Tauri least-privilege capabilities + strict CSP + minisign-signed updater (any manifest/key compromise = full takeover). https://v2.tauri.app/security/
- CWE Top 25 (2025-12-11) — #1 XSS, #2 SQLi, #3 CSRF; watch CWE-284 (Improper Access Control), CWE-639 (authz bypass via user-controlled key), CWE-770 (alloc without limits). https://www.cisa.gov/news-events/alerts/2025/12/11/2025-cwe-top-25-most-dangerous-software-weaknesses
Common mistakes: citing Top 10 2021 / ASVS 4.x / SLSA v1.0 (all superseded); treating npm/SLSA provenance as a safety guarantee (Shai-Hulud defeated valid L3); wildcard Tauri capabilities or no CSP ("it's a desktop app"); secrets in tauri.conf.json/env/logs; lockfile uncommitted or CI without --frozen-lockfile; no cargo audit/cargo deny gate; ignoring A03/A10 + LLM01/LLM06 for AI features.
