security-review
/security-review
Ocenia funkcję lub PR pod kątem bezpieczeństwa i lekkiego modelu zagrożeń: uwierzytelnianie, dostęp do danych, wstrzyknięcia, sekrety i logowanie.
- Autor
- omexit ↗
- Repozytorium
- claude-skills-pack ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 15.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 2
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 15.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Ocenia funkcję lub PR pod kątem bezpieczeństwa i lekkiego modelu zagrożeń: uwierzytelnianie, dostęp do danych, wstrzyknięcia, sekrety i logowanie.
Oryginalny opis (EN)
Use when a feature or PR needs a security review or lightweight threat model: auth, data access, injection, abuse cases, privacy, secrets, logging, or is this safe.
Jak działa
- Inputs (ask only if missing)
- Lenses and fan-out
- Workflow
- Output
- Workflow context
- Output contract
Podgląd SKILL.md
---
name: security-review
description: "Use when a feature or PR needs a security review or lightweight threat model: auth, data access, injection, abuse cases, privacy, secrets, logging, or is this safe."
---
# Security review (practical)
> Delegation, model labels, and task tracking below use Claude Code notation. In Codex or another client, follow `../../references/cross-client.md`.
A lightweight threat model and findings for one feature, PR, or endpoint set, prioritized by real
exploitability. This is the security deep dive; `code-audit` covers security only as one breadth
dimension, and `arch-review` covers structural boundaries.
## Inputs (ask only if missing)
- What is changing (feature, PR, endpoints)
- Data involved (PII, financial, secrets) and the auth model (who can do what)
- Deployment context (public, internal, multi-tenant)
## Lenses and fan-out
| Lens | Label | Focus |
|---|---|---|
| `AUTH` | `opus` | Authentication, session, server-side authorization, object-level access, tenant isolation, roles |
| `INJECTION` | `sonnet` | SQL/NoSQL, command, template, path traversal, XSS, SSRF, unsafe deserialization |
| `PRIVACY` | `opus` | PII minimization, retention, logs and errors, analytics payloads, secrets handling |
| `ABUSE` | `sonnet` | Rate limits, enumeration, credential stuffing, resource exhaustion, bulk operations |
Map the attack surface first, in one pass. **Fan-out is conditional:** a single endpoint, PR, or
module runs the four lenses inline. Fan out only for a large surface (many services or entry
points): dispatch the lenses as parallel Agent calls in one message with the attack-surface map,
then synthesize one risk-ranked list.
## Workflow
1. **Attack surface:** every entry point (endpoints, forms, uploads, webhooks, consumers, jobs,
headers, URL parameters) and the trust boundary each crosses.
2. **Trace data flow:** for each input, where it is stored, rendered, logged, and forwarded.
3. **Authentication and authorization:** per endpoint and per persistence path, enforced
server-side; object-level checks (can user A reach user B's record by changing an ID?); tenant
isolation; privilege escalation between roles; default deny.
4. **Injection and parsing:** parameterized queries, escaped output, no user input in shell,
templates, file paths, or fetched URLs without validation.
5. **Exposure:** response over-fetching; PII, tokens, or secrets in logs, errors, analytics, or URLs;
errors returned as RFC 9457 Problem Details without stack traces or provider internals; internal
IDs (TSIDs) never exposed, external IDs are UUIDv7.
6. **Secrets and dependencies:** no secrets in code, committed config, logs, or URLs; secrets come
from the platform's secret store or injected environment. Check vulnerable dependencies against
the resolved dependency graph with official advisories or the ecosystem scanner; never infer a
vulnerability from a package name or a different version.
7. **Abuse:** enumeration, credential stuffing, spam through user content, unbounded bulk requests,
expensive operations without limits.
8. **Defaults:** new features fail closed; debug endpoints, permissive CORS, and default
credentials are off in production.
Every finding states what is vulnerable, the exploit path (who, from where, with what access), the
evidence (code, config, request, or scoped search), and the fix. Theoretical risk with no reachable
path is a note, not a finding. Never propose home-grown cryptography.
## Output
Fill `assets/security-review.md`. Group findings as **Must-fix** (exploitable now, high impact),
**Should-fix** (exploitable with effort, moderate impact), **Nice-to-have** (defense in depth).
State explicitly what was not reviewed, and give a verification step for each fix.
## Workflow context
- Typically follows: `design-doc`, `code-audit` (security leads), `code-review:code-review` (official plugin).
- Feeds into: `test-plan` (security test cases), `finalize`.
- Related: `performance-review` (rate limiting and resource exhaustion overlap).
## Output contract
```yaml
produces:
- type: "security-review"
format: "markdown"
path: "claudedocs/<feature>-security-review.md"
sections: [threat_model, findings, owasp_checklist, remediation]
handoff: "Write claudedocs/handoff-security-review-<timestamp>.yaml - suggest: finalize, test-plan"
```
Pliki w paczce (2)
- SKILL.md 4 KB
- assets/security-review.md 2 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · omexit/claude-skills-pack | 100 | MIT |
| microsoft/power-platform-skills | 100 | MIT |
| VasulenkoIllia/claude-code-base | 100 | MIT |
| rjmurillo/ai-agents | 100 | MIT |
| Sanit-mann/The-Insectoids | 100 | MIT |
| github/awesome-copilot | 100 | MIT |
| sageox/ox | 100 | MIT |
| mcouthon/agents | 100 | MIT |
| relatedb/relate | 100 | Apache-2.0 |
Migawka bazy z 13.11.2026 · wersja paczki c792da707aa6 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 95security-privacy-threat-modeling
Prowadzi modelowanie zagrożeń, przegląd bezpiecznego kodu i privacy by design, z bramkami akceptacji bezpieczeństwa. - 100security-audit
Audytuje zmianę na podstawie modelu zagrożeń: kontrolę dostępu, wstrzyknięcia, sekrety i dane wrażliwe w diffie lub PR. - 90security-threat-model-and-design
Ocenia i naprawia modele zagrożeń, scenariusze nadużyć, architekturę i przeglądy bezpiecznego projektu we frontendzie oraz usługach. - 100security-review-spec
Sprawdza pull request ze specyfikacją produktu lub technicznej pod kątem powierzchni zagrożeń, modelu uwierzytelniania i wrażliwych danych. - 100security-audit
Przeprowadza szybki przegląd bezpieczeństwa przed pull requestem lub pełny audyt repozytorium: sekrety, zależności i logowanie. - 90security-audit
Przegląda całe repozytorium pod kątem bezpieczeństwa: kod, konfigurację, IaC, CI/CD, zależności i sekrety, z modelem zagrożeń i planem napraw.