ClawHub
Back to skill

Security audit

Split

Security checks for vulnerabilities and agentic risk

Overview

This skill is a static reference for data-splitting techniques and its bundled script only prints educational guidance.

Installers should understand this is an educational command-reference skill. Some displayed examples, if manually copied and run by a user, can create chunk files or process data, so users should apply them only to intended files and review paths before running examples themselves.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

YARA rule 'agent_skill_mcp_tool_poisoning_metadata': MCP/tool metadata poisoning indicators in tool schemas or skill manifests [agent_skills]

High
Category
YARA Match
Confidence
80% confidence
Finding

YARA rule matched a hack tool or exploit indicator (offensive tools, reconnaissance, privilege escalation, or exploit frameworks).

Content

Scanner excerpt · scripts/script.sh (reported line 131)May include surrounding context.

sh
print > ("shard_" NR%4 ".txt")}' input.txt
  Result: 4 files with interleaved lines
  Use: Load balancing, parallel worker distribution

Preserving Headers:
  # Split CSV keeping header in each chunk
  head -1 data.csv > header.txt
  tail -n +2 data.csv | split -l 10000 - chunk_
  for f in chunk_*; do cat header.txt "$f" > "h_$f"; done
  Critical for: CSV/TSV files processed independently

Common Tools:
  split      GNU coreutils — split by size or lines
  csplit     Context-aware split by pattern
  awk        Flexible splitting with custom logic
  zipsplit   Split zip archives
  tar        Split with: tar cf - dir | split -b 1G - archive_
  ffmpeg     Split video: ffmpeg -ss 00:05:00 -t 00:10:00

Reassembly:
  cat part_* > reassembled.bin
  Verify: md5sum original.bin reassembled.bin
EOF
}

cmd_dataset() {
    cat << 'EOF'
=== ML Dataset Splitting ===

Standard Split (Train/Validation/Test):
  Train       60-80%  Model learns from this data
  Validation  10-20%  Tune hyperparameters

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The description says to use the skill when 'dividing data, chunking files, or designing data partitioning strategies,' but it does not define explicit trigger phrases, constraints, or non-matching examples. Terms like 'chunking files' and 'dividing data' are broad enough to match many ordinary requests, which can cause unintended activation in systems that rely on natural-language descriptions.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.