Skip to content

Commit 3d8366c

Browse files
ericapisaniclaude
ericapisani
and
claude
committed
chore(agents): Add security-review skill to agent configuration
Enable the security-review skill in Warden to provide AI-powered security analysis on pull requests. The skill reviews code for vulnerabilities and security issues, triggered on PR open and updates. Configures the skill in agents.toml, agents.lock, and warden.toml with triggers for opened, synchronize, and reopened actions. Co-Authored-By: Claude <noreply@anthropic.com>
1 parent f08ab01 commit 3d8366c

25 files changed

+8388
-0
lines changed

‎.agents/skills/security-review/LICENSE‎

Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
The reference material in this skill is derived from the OWASP Cheat Sheet Series.
2+
3+
Source: https://cheatsheetseries.owasp.org/
4+
OWASP Foundation: https://owasp.org/
5+
6+
Original content is licensed under:
7+
8+
Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)
9+
https://creativecommons.org/licenses/by-sa/4.0/
10+
11+
You are free to:
12+
- Share — copy and redistribute the material in any medium or format
13+
- Adapt — remix, transform, and build upon the material for any purpose,
14+
even commercially
15+
16+
Under the following terms:
17+
- Attribution — You must give appropriate credit, provide a link to the
18+
license, and indicate if changes were made.
19+
- ShareAlike — If you remix, transform, or build upon the material, you
20+
must distribute your contributions under the same license as the original.
21+
22+
Full license text: https://creativecommons.org/licenses/by-sa/4.0/legalcode

‎.agents/skills/security-review/SKILL.md‎

Lines changed: 312 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,312 @@
1+
---
2+
name: security-review
3+
description: Security code review for vulnerabilities. Use when asked to "security review", "find vulnerabilities", "check for security issues", "audit security", "OWASP review", or review code for injection, XSS, authentication, authorization, cryptography issues. Provides systematic review with confidence-based reporting.
4+
allowed-tools: Read, Grep, Glob, Bash, Task
5+
license: LICENSE
6+
---
7+
8+
<!--
9+
Reference material based on OWASP Cheat Sheet Series (CC BY-SA 4.0)
10+
https://cheatsheetseries.owasp.org/
11+
-->
12+
13+
# Security Review Skill
14+
15+
Identify exploitable security vulnerabilities in code. Report only **HIGH CONFIDENCE** findings—clear vulnerable patterns with attacker-controlled input.
16+
17+
## Scope: Research vs. Reporting
18+
19+
**CRITICAL DISTINCTION:**
20+
21+
- **Report on**: Only the specific file, diff, or code provided by the user
22+
- **Research**: The ENTIRE codebase to build confidence before reporting
23+
24+
Before flagging any issue, you MUST research the codebase to understand:
25+
- Where does this input actually come from? (Trace data flow)
26+
- Is there validation/sanitization elsewhere?
27+
- How is this configured? (Check settings, config files, middleware)
28+
- What framework protections exist?
29+
30+
**Do NOT report issues based solely on pattern matching.** Investigate first, then report only what you're confident is exploitable.
31+
32+
## Confidence Levels
33+
34+
| Level | Criteria | Action |
35+
|-------|----------|--------|
36+
| **HIGH** | Vulnerable pattern + attacker-controlled input confirmed | **Report** with severity |
37+
| **MEDIUM** | Vulnerable pattern, input source unclear | **Note** as "Needs verification" |
38+
| **LOW** | Theoretical, best practice, defense-in-depth | **Do not report** |
39+
40+
## Do Not Flag
41+
42+
### General Rules
43+
- Test files (unless explicitly reviewing test security)
44+
- Dead code, commented code, documentation strings
45+
- Patterns using **constants** or **server-controlled configuration**
46+
- Code paths that require prior authentication to reach (note the auth requirement instead)
47+
48+
### Server-Controlled Values (NOT Attacker-Controlled)
49+
50+
These are configured by operators, not controlled by attackers:
51+
52+
| Source | Example | Why It's Safe |
53+
|--------|---------|---------------|
54+
| Django settings | `settings.API_URL`, `settings.ALLOWED_HOSTS` | Set via config/env at deployment |
55+
| Environment variables | `os.environ.get('DATABASE_URL')` | Deployment configuration |
56+
| Config files | `config.yaml`, `app.config['KEY']` | Server-side files |
57+
| Framework constants | `django.conf.settings.*` | Not user-modifiable |
58+
| Hardcoded values | `BASE_URL = "https://api.internal"` | Compile-time constants |
59+
60+
**SSRF Example - NOT a vulnerability:**
61+
```python
62+
# SAFE: URL comes from Django settings (server-controlled)
63+
response = requests.get(f"{settings.SEER_AUTOFIX_URL}{path}")
64+
```
65+
66+
**SSRF Example - IS a vulnerability:**
67+
```python
68+
# VULNERABLE: URL comes from request (attacker-controlled)
69+
response = requests.get(request.GET.get('url'))
70+
```
71+
72+
### Framework-Mitigated Patterns
73+
Check language guides before flagging. Common false positives:
74+
75+
| Pattern | Why It's Usually Safe |
76+
|---------|----------------------|
77+
| Django `{{ variable }}` | Auto-escaped by default |
78+
| React `{variable}` | Auto-escaped by default |
79+
| Vue `{{ variable }}` | Auto-escaped by default |
80+
| `User.objects.filter(id=input)` | ORM parameterizes queries |
81+
| `cursor.execute("...%s", (input,))` | Parameterized query |
82+
| `innerHTML = "<b>Loading...</b>"` | Constant string, no user input |
83+
84+
**Only flag these when:**
85+
- Django: `{{ var|safe }}`, `{% autoescape off %}`, `mark_safe(user_input)`
86+
- React: `dangerouslySetInnerHTML={{__html: userInput}}`
87+
- Vue: `v-html="userInput"`
88+
- ORM: `.raw()`, `.extra()`, `RawSQL()` with string interpolation
89+
90+
## Review Process
91+
92+
### 1. Detect Context
93+
94+
What type of code am I reviewing?
95+
96+
| Code Type | Load These References |
97+
|-----------|----------------------|
98+
| API endpoints, routes | `authorization.md`, `authentication.md`, `injection.md` |
99+
| Frontend, templates | `xss.md`, `csrf.md` |
100+
| File handling, uploads | `file-security.md` |
101+
| Crypto, secrets, tokens | `cryptography.md`, `data-protection.md` |
102+
| Data serialization | `deserialization.md` |
103+
| External requests | `ssrf.md` |
104+
| Business workflows | `business-logic.md` |
105+
| GraphQL, REST design | `api-security.md` |
106+
| Config, headers, CORS | `misconfiguration.md` |
107+
| CI/CD, dependencies | `supply-chain.md` |
108+
| Error handling | `error-handling.md` |
109+
| Audit, logging | `logging.md` |
110+
111+
### 2. Load Language Guide
112+
113+
Based on file extension or imports:
114+
115+
| Indicators | Guide |
116+
|------------|-------|
117+
| `.py`, `django`, `flask`, `fastapi` | `languages/python.md` |
118+
| `.js`, `.ts`, `express`, `react`, `vue`, `next` | `languages/javascript.md` |
119+
| `.go`, `go.mod` | `languages/go.md` |
120+
| `.rs`, `Cargo.toml` | `languages/rust.md` |
121+
| `.java`, `spring`, `@Controller` | `languages/java.md` |
122+
123+
### 3. Load Infrastructure Guide (if applicable)
124+
125+
| File Type | Guide |
126+
|-----------|-------|
127+
| `Dockerfile`, `.dockerignore` | `infrastructure/docker.md` |
128+
| K8s manifests, Helm charts | `infrastructure/kubernetes.md` |
129+
| `.tf`, Terraform | `infrastructure/terraform.md` |
130+
| GitHub Actions, `.gitlab-ci.yml` | `infrastructure/ci-cd.md` |
131+
| AWS/GCP/Azure configs, IAM | `infrastructure/cloud.md` |
132+
133+
### 4. Research Before Flagging
134+
135+
**For each potential issue, research the codebase to build confidence:**
136+
137+
- Where does this value actually come from? Trace the data flow.
138+
- Is it configured at deployment (settings, env vars) or from user input?
139+
- Is there validation, sanitization, or allowlisting elsewhere?
140+
- What framework protections apply?
141+
142+
Only report issues where you have HIGH confidence after understanding the broader context.
143+
144+
### 5. Verify Exploitability
145+
146+
For each potential finding, confirm:
147+
148+
**Is the input attacker-controlled?**
149+
150+
| Attacker-Controlled (Investigate) | Server-Controlled (Usually Safe) |
151+
|-----------------------------------|----------------------------------|
152+
| `request.GET`, `request.POST`, `request.args` | `settings.X`, `app.config['X']` |
153+
| `request.json`, `request.data`, `request.body` | `os.environ.get('X')` |
154+
| `request.headers` (most headers) | Hardcoded constants |
155+
| `request.cookies` (unsigned) | Internal service URLs from config |
156+
| URL path segments: `/users/<id>/` | Database content from admin/system |
157+
| File uploads (content and names) | Signed session data |
158+
| Database content from other users | Framework settings |
159+
| WebSocket messages | |
160+
161+
**Does the framework mitigate this?**
162+
- Check language guide for auto-escaping, parameterization
163+
- Check for middleware/decorators that sanitize
164+
165+
**Is there validation upstream?**
166+
- Input validation before this code
167+
- Sanitization libraries (DOMPurify, bleach, etc.)
168+
169+
### 6. Report HIGH Confidence Only
170+
171+
Skip theoretical issues. Report only what you've confirmed is exploitable after research.
172+
173+
---
174+
175+
## Severity Classification
176+
177+
| Severity | Impact | Examples |
178+
|----------|--------|----------|
179+
| **Critical** | Direct exploit, severe impact, no auth required | RCE, SQL injection to data, auth bypass, hardcoded secrets |
180+
| **High** | Exploitable with conditions, significant impact | Stored XSS, SSRF to metadata, IDOR to sensitive data |
181+
| **Medium** | Specific conditions required, moderate impact | Reflected XSS, CSRF on state-changing actions, path traversal |
182+
| **Low** | Defense-in-depth, minimal direct impact | Missing headers, verbose errors, weak algorithms in non-critical context |
183+
184+
---
185+
186+
## Quick Patterns Reference
187+
188+
### Always Flag (Critical)
189+
```
190+
eval(user_input) # Any language
191+
exec(user_input) # Any language
192+
pickle.loads(user_data) # Python
193+
yaml.load(user_data) # Python (not safe_load)
194+
unserialize($user_data) # PHP
195+
deserialize(user_data) # Java ObjectInputStream
196+
shell=True + user_input # Python subprocess
197+
child_process.exec(user) # Node.js
198+
```
199+
200+
### Always Flag (High)
201+
```
202+
innerHTML = userInput # DOM XSS
203+
dangerouslySetInnerHTML={user} # React XSS
204+
v-html="userInput" # Vue XSS
205+
f"SELECT * FROM x WHERE {user}" # SQL injection
206+
`SELECT * FROM x WHERE ${user}` # SQL injection
207+
os.system(f"cmd {user_input}") # Command injection
208+
```
209+
210+
### Always Flag (Secrets)
211+
```
212+
password = "hardcoded"
213+
api_key = "sk-..."
214+
AWS_SECRET_ACCESS_KEY = "..."
215+
private_key = "-----BEGIN"
216+
```
217+
218+
### Check Context First (MUST Investigate Before Flagging)
219+
```
220+
# SSRF - ONLY if URL is from user input, NOT from settings/config
221+
requests.get(request.GET['url']) # FLAG: User-controlled URL
222+
requests.get(settings.API_URL) # SAFE: Server-controlled config
223+
requests.get(f"{settings.BASE}/{x}") # CHECK: Is 'x' user input?
224+
225+
# Path traversal - ONLY if path is from user input
226+
open(request.GET['file']) # FLAG: User-controlled path
227+
open(settings.LOG_PATH) # SAFE: Server-controlled config
228+
open(f"{BASE_DIR}/{filename}") # CHECK: Is 'filename' user input?
229+
230+
# Open redirect - ONLY if URL is from user input
231+
redirect(request.GET['next']) # FLAG: User-controlled redirect
232+
redirect(settings.LOGIN_URL) # SAFE: Server-controlled config
233+
234+
# Weak crypto - ONLY if used for security purposes
235+
hashlib.md5(file_content) # SAFE: File checksums, caching
236+
hashlib.md5(password) # FLAG: Password hashing
237+
random.random() # SAFE: Non-security uses (UI, sampling)
238+
random.random() for token # FLAG: Security tokens need secrets module
239+
```
240+
241+
---
242+
243+
## Output Format
244+
245+
```markdown
246+
## Security Review: [File/Component Name]
247+
248+
### Summary
249+
- **Findings**: X (Y Critical, Z High, ...)
250+
- **Risk Level**: Critical/High/Medium/Low
251+
- **Confidence**: High/Mixed
252+
253+
### Findings
254+
255+
#### [VULN-001] [Vulnerability Type] (Severity)
256+
- **Location**: `file.py:123`
257+
- **Confidence**: High
258+
- **Issue**: [What the vulnerability is]
259+
- **Impact**: [What an attacker could do]
260+
- **Evidence**:
261+
```python
262+
[Vulnerable code snippet]
263+
```
264+
- **Fix**: [How to remediate]
265+
266+
### Needs Verification
267+
268+
#### [VERIFY-001] [Potential Issue]
269+
- **Location**: `file.py:456`
270+
- **Question**: [What needs to be verified]
271+
```
272+
273+
If no vulnerabilities found, state: "No high-confidence vulnerabilities identified."
274+
275+
---
276+
277+
## Reference Files
278+
279+
### Core Vulnerabilities (`references/`)
280+
| File | Covers |
281+
|------|--------|
282+
| `injection.md` | SQL, NoSQL, OS command, LDAP, template injection |
283+
| `xss.md` | Reflected, stored, DOM-based XSS |
284+
| `authorization.md` | Authorization, IDOR, privilege escalation |
285+
| `authentication.md` | Sessions, credentials, password storage |
286+
| `cryptography.md` | Algorithms, key management, randomness |
287+
| `deserialization.md` | Pickle, YAML, Java, PHP deserialization |
288+
| `file-security.md` | Path traversal, uploads, XXE |
289+
| `ssrf.md` | Server-side request forgery |
290+
| `csrf.md` | Cross-site request forgery |
291+
| `data-protection.md` | Secrets exposure, PII, logging |
292+
| `api-security.md` | REST, GraphQL, mass assignment |
293+
| `business-logic.md` | Race conditions, workflow bypass |
294+
| `modern-threats.md` | Prototype pollution, LLM injection, WebSocket |
295+
| `misconfiguration.md` | Headers, CORS, debug mode, defaults |
296+
| `error-handling.md` | Fail-open, information disclosure |
297+
| `supply-chain.md` | Dependencies, build security |
298+
| `logging.md` | Audit failures, log injection |
299+
300+
### Language Guides (`languages/`)
301+
- `python.md` - Django, Flask, FastAPI patterns
302+
- `javascript.md` - Node, Express, React, Vue, Next.js
303+
- `go.md` - Go-specific security patterns
304+
- `rust.md` - Rust unsafe blocks, FFI security
305+
- `java.md` - Spring, Java EE patterns
306+
307+
### Infrastructure (`infrastructure/`)
308+
- `docker.md` - Container security
309+
- `kubernetes.md` - K8s RBAC, secrets, policies
310+
- `terraform.md` - IaC security
311+
- `ci-cd.md` - Pipeline security
312+
- `cloud.md` - AWS/GCP/Azure security

0 commit comments

Comments
 (0)