Liberty91LTD/cti-skills / skills/yara-writing
yara-writing
VerifiedUse when the user asks for a YARA rule, "write YARA for this sample/family", or `/hash-investigation` / `/malware-analysis` surfaces a sample worth a static-content rule. Pattern-matching rules for identifying malicious files.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 6.1 KB |
SKILL.md
View on GitHub---
name: yara-writing
description: Use when the user asks for a YARA rule, "write YARA for this sample/family", or `/hash-investigation` / `/malware-analysis` surfaces a sample worth a static-content rule. Pattern-matching rules for identifying malicious files.
user-invocable: true
metadata:
version: 1.1.0
---
# YARA Rule Writing Guide
YARA rules identify and classify malware by matching byte patterns, strings, and conditions in files.
## Rule Structure
```yara
rule RuleName : tag1 tag2 {
meta:
author = "CTI Platform"
date = "YYYY-MM-DD"
description = "Description of what this rule detects"
reference = "https://source-reference.com"
tlp = "GREEN"
mitre_attack = "T1566"
hash = "sha256_of_sample"
confidence = "high"
strings:
$text1 = "suspicious string" ascii wide
$hex1 = { 4D 5A 90 00 03 00 00 00 }
$regex1 = /https?:\/\/[a-z0-9\-\.]+\.onion/
condition:
uint16(0) == 0x5A4D and
filesize < 5MB and
(2 of ($text*) or $hex1) and
$regex1
}
```
## String Types
### Text strings
```yara
$s1 = "CreateRemoteThread" # Exact ASCII match
$s2 = "kernel32.dll" ascii wide # Match both ASCII and UTF-16
$s3 = "password" nocase # Case-insensitive
$s4 = "cmd /c" ascii wide nocase # Combined modifiers
$s5 = "C:\\Windows\\Temp\\" # Escaped backslashes
```
### Hex strings
```yara
$h1 = { 4D 5A 90 00 } # Exact bytes (PE header)
$h2 = { 4D 5A ?? ?? 03 00 } # Wildcards (?? = any byte)
$h3 = { 4D 5A [2-4] 03 00 } # Jump (2-4 bytes between)
$h4 = { 4D 5A ( 90 00 | 00 00 ) } # Alternation (OR)
```
### Regular expressions
```yara
$r1 = /https?:\/\/[\w\-\.]+/ # URL pattern
$r2 = /[A-Za-z0-9+\/]{50,}={0,2}/ # Base64 encoded content
$r3 = /(\d{1,3}\.){3}\d{1,3}/ # IP address pattern
```
## Conditions
### File properties
```yara
uint16(0) == 0x5A4D # PE file (MZ header)
uint32(0) == 0x464C457F # ELF file
uint32(0) == 0xBEBAFECA # Mach-O fat binary
filesize < 10MB # File size limit
```
### String matching
```yara
all of them # All strings must match
any of them # At least one string
2 of ($text*) # At least 2 of $text group
3 of ($s1, $s2, $s3, $s4) # 3 of these 4 strings
$s1 at 0 # $s1 at offset 0
$s1 in (0..1024) # $s1 in first 1KB
#s1 > 3 # $s1 occurs more than 3 times
```
### PE module (Windows executables)
```yara
import "pe"
pe.number_of_sections > 5
pe.imports("kernel32.dll", "VirtualAlloc")
pe.exports("DllRegisterServer")
pe.is_dll()
pe.timestamp > 1704067200 # Compiled after 2024-01-01
```
## Best Practices
### Do
- Start with a PE/ELF/file type check to limit scope
- Add a filesize constraint to avoid scanning large files
- Use unique strings — prefer byte sequences or specific API combinations over generic strings
- Combine multiple indicators (string + structure + behavior)
- Include metadata with reference to source intelligence
- Test against both malicious samples AND clean files (false positive testing)
### Don't
- Match on single, common strings (e.g., just "http://" or "cmd.exe")
- Create rules without file type constraints (performance killer)
- Use overly broad regex patterns
- Forget to escape special characters in hex patterns
- Write rules that match on packer/crypter artifacts only (will match unrelated packed software)
## Common Malware Patterns
### Encoded PowerShell
```yara
rule Encoded_PowerShell_Execution {
meta:
description = "Detects base64 encoded PowerShell commands"
mitre_attack = "T1059.001"
strings:
$enc1 = "-EncodedCommand" ascii nocase
$enc2 = "-enc " ascii nocase
$enc3 = "FromBase64String" ascii nocase
$b64 = /[A-Za-z0-9+\/]{100,}={0,2}/ ascii
condition:
any of ($enc*) and $b64
}
```
### Suspicious PE imports
```yara
import "pe"
rule Suspicious_Injection_Imports {
meta:
description = "PE with process injection API imports"
mitre_attack = "T1055"
condition:
pe.imports("kernel32.dll", "VirtualAllocEx") and
pe.imports("kernel32.dll", "WriteProcessMemory") and
pe.imports("kernel32.dll", "CreateRemoteThread")
}
```
## Naming Convention
- `APT_<GroupName>_<MalwareName>.yar` — For attributed malware
- `MAL_<MalwareName>_<Variant>.yar` — For named malware families
- `SUSP_<Behavior>_<Context>.yar` — For suspicious behavior patterns
- `HUNT_<Target>_<Technique>.yar` — For threat hunting queries
## Output Location
Write YARA rules to: `data/detection-rules/yara/<malware-name>.yar`
## Validation against a real corpus
Before deploying a rule beyond your test set, validate it against a broader sample collection. When ReversingLabs Spectra Analyze is available:
- `/lookup-reversinglabs yara-matches <ruleset_name>` — for a rule you have already deployed to RL, list the samples it has matched in the continuous-scanning corpus. Useful to confirm the rule isn't firing on goodware and to discover new variants caught by the same logic.
- `/lookup-reversinglabs search 'threatname:<family> classification:malicious' --max-results 100` — before writing a family rule, retrieve a candidate sample set to test against. Then pull each sample's hashes for inclusion in your test suite.
- `/lookup-reversinglabs containers <hash>` and `extracted <hash>` — find parents/children of a known sample to ensure the rule fires across the dropper-payload chain, not just on one layer.
Without RL access, fall back to retro-hunting via `/lookup-virustotal` (premium) and `/lookup-otx` pulse traversal.
## Related skills
- `/hash-investigation`, `/malware-analysis` — produce the structural and behavioural intel that informs the rule
- `/lookup-reversinglabs` — corpus validation, family pivots, parent/child relationships
- `/lookup-virustotal`, `/lookup-otx` — community detection signals and pulse-driven sample collection
- `/sigma-writing` — behavioural-detection counterpart for log telemetry
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |