Gdpr Security Auditor

• Outcome: Review
• Latest audit: Jul 4, 2026 · 9:01 PM UTC
• Version: 1.0.0

Overview

This GDPR audit skill is mostly purpose-aligned, but it asks an agent to perform broad sensitive system scans and privileged checks without tight scope or output protections.

Install or run this only on systems you are authorized to audit. Limit scans to approved directories and accounts, run privileged checks only when needed, and protect the generated report directory with restrictive permissions because it may contain PII locations, credential file paths, access-control details, and security weaknesses.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (14)

Intent-Code Divergence

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly claims the audit is "read-only," but the documented workflows create directories and write multiple audit artifacts and templates to disk via mkdir, tee redirections, and here-doc output. This mismatch is dangerous because operators may grant access or run the skill under a false assumption about side effects, which can affect integrity, storage handling, and retention of sensitive audit data.

External Transmission

Medium

• Category: Data Exfiltration

Content

echo "--- TLS Verification ---"
for port in 443 8443 9443; do
  echo "Port $port:"
  result=$(curl -s -X POST "http://localhost:8765/aynops/ssl" \
    -H "Content-Type: application/json" \
    -d "{\"host\": \"$TARGET\", \"port\": $port}" 2>/dev/null)

• Confidence: 86% confidence
• Finding: The skill transmits target host and port information over HTTP to a local service endpoint for TLS analysis. Even though the endpoint is localhost, this creates a data flow to another component that may log, mishandle, or forward audit metadata, and the use of unauthenticated plain HTTP weakens the trust boundary.

File System Enumeration

Medium

• Category: Data Exfiltration

Content

# World-readable sensitive files
echo ""
echo "--- World-Readable Sensitive Files ---"
find /home -type f \( -name "*.env" -o -name "*.pem" -o -name "id_*" -o -name "*.key" \) -perm /o+r 2>/dev/null | \
  while IFS= read -r file; do
    echo "  ❌ WORLD-READABLE: $file"
  done | tee "$AUDIT_DIR/world-readable-sensitive.txt"

• Confidence: 91% confidence
• Finding: The skill performs broad filesystem enumeration under /home for sensitive file types, which can expose the presence and paths of secrets, private keys, and credentials across user accounts. In an audit context this may be intentional, but it still increases access to sensitive metadata and can centralize high-value information into report files.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

# PostgreSQL
if systemctl is-active --quiet postgresql 2>/dev/null; then
  echo "PostgreSQL: RUNNING" | tee -a "$AUDIT_DIR/databases.txt"
  sudo -u postgres psql -c "\l" 2>/dev/null >> "$AUDIT_DIR/postgres-databases.txt"
else
  echo "PostgreSQL: not running" | tee -a "$AUDIT_DIR/databases.txt"
fi

• Confidence: 90% confidence
• Finding: The skill invokes sudo to access PostgreSQL inventory data as the postgres user, requiring elevated privileges beyond a normal unprivileged audit. If the skill is run broadly or by automation, this expands the blast radius and may expose database names and metadata that should be tightly controlled.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

# Firewall rules
echo ""
echo "--- Firewall Rules ---"
sudo iptables -L -n -v 2>/dev/null | head -40 | tee "$AUDIT_DIR/firewall-rules.txt"

# Open ports
echo ""

• Confidence: 93% confidence
• Finding: The skill runs sudo iptables to enumerate firewall rules, requiring elevated privileges. In the context of an agent skill, any use of sudo is security-relevant because it conditions operators to grant privilege and may reveal sensitive network policy details while increasing the consequences of misuse.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

# PostgreSQL table sizes and ages
if systemctl is-active --quiet postgresql 2>/dev/null; then
  sudo -u postgres psql -c "
    SELECT 
      schemaname || '.' || relname AS table_name,
      n_live_tup AS estimated_rows,

• Confidence: 90% confidence
• Finding: This command uses sudo -u postgres to query PostgreSQL table metadata, again requiring elevation into a privileged service account. While operationally useful for retention analysis, it broadens data access and can normalize running the skill with more privilege than necessary.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

| Symptom | Likely Cause | Fix |
|---------|-------------|-----|
| Disk encryption not detected | Not using LUKS | Check with `sudo dmsetup ls` |
| TLS check fails | Certificate expired | Renew via certbot or CA |
| Firewall rules empty | iptables not configured | `sudo iptables -L -n` |
| PostgreSQL access denied | No sudo/peer auth | Configure `~/.pgpass` or peer auth |

• Confidence: 78% confidence
• Finding: The troubleshooting section recommends running sudo dmsetup ls, which encourages operators to use elevated privileges during troubleshooting. This is lower risk than embedded execution, but still broadens the expected privilege model of the skill and may lead to unnecessary root use.

Sudo/Root Execution

Medium

• Category: Privilege Escalation

Content

|---------|-------------|-----|
| Disk encryption not detected | Not using LUKS | Check with `sudo dmsetup ls` |
| TLS check fails | Certificate expired | Renew via certbot or CA |
| Firewall rules empty | iptables not configured | `sudo iptables -L -n` |
| PostgreSQL access denied | No sudo/peer auth | Configure `~/.pgpass` or peer auth |
| PII scan too slow on large filesystems | Recursive find on /home | Limit with `-maxdepth 3` |
| DPIA template questions unclear | Template designed for general case | Customize per processing activity |

• Confidence: 78% confidence
• Finding: The documentation recommends sudo iptables -L -n for troubleshooting, which normalizes elevated execution. While not automatically executed by the skill, it still pushes operators toward broader privilege use than may be necessary.

Credential Access

High

• Category: Privilege Escalation

Content

echo "--- SSH Keys ---"
for user_home in /home/*; do
  user=$(basename "$user_home")
  if [ -f "$user_home/.ssh/authorized_keys" ]; then
    key_count=$(wc -l < "$user_home/.ssh/authorized_keys")
    echo "  User $user: $key_count authorized keys"

• Confidence: 94% confidence
• Finding: The skill enumerates users' authorized_keys files, which are sensitive authentication artifacts. Even counting and classifying key types across user homes reveals credential deployment patterns and, when centralized in reports, creates a high-value map of access paths for an attacker or over-privileged operator.

Credential Access

High

• Category: Privilege Escalation

Content

for user_home in /home/*; do
  user=$(basename "$user_home")
  if [ -f "$user_home/.ssh/authorized_keys" ]; then
    key_count=$(wc -l < "$user_home/.ssh/authorized_keys")
    echo "  User $user: $key_count authorized keys"
    
    # Check key types (ed25519 > rsa)

• Confidence: 94% confidence
• Finding: This access to ~/.ssh/authorized_keys continues the same credential-sensitive enumeration and records counts per user. In a multi-user environment, this can disclose which accounts are remotely accessible and how broadly keys are deployed.

Credential Access

High

• Category: Privilege Escalation

Content

echo "  User $user: $key_count authorized keys"
    
    # Check key types (ed25519 > rsa)
    rsa_keys=$(grep -c "ssh-rsa" "$user_home/.ssh/authorized_keys" 2>/dev/null || echo 0)
    ed25519_keys=$(grep -c "ssh-ed25519" "$user_home/.ssh/authorized_keys" 2>/dev/null || echo 0)
    if [ "$rsa_keys" -gt 0 ]; then
      echo "    ⚠️  $rsa_keys RSA keys (deprecated — migrate to ed25519)"

• Confidence: 95% confidence
• Finding: Grepping authorized_keys for key types inspects authentication material and reveals cryptographic posture of user access. This is legitimate for an audit, but it still touches sensitive credential locations and aggregates information that could aid targeted intrusion planning.

Credential Access

High

• Category: Privilege Escalation

Content

# Check key types (ed25519 > rsa)
    rsa_keys=$(grep -c "ssh-rsa" "$user_home/.ssh/authorized_keys" 2>/dev/null || echo 0)
    ed25519_keys=$(grep -c "ssh-ed25519" "$user_home/.ssh/authorized_keys" 2>/dev/null || echo 0)
    if [ "$rsa_keys" -gt 0 ]; then
      echo "    ⚠️  $rsa_keys RSA keys (deprecated — migrate to ed25519)"
    fi

• Confidence: 95% confidence
• Finding: The continued inspection of authorized_keys compounds the credential-access concern by identifying deprecated RSA key usage per account. This information is valuable for remediation, but also valuable to an attacker seeking weaker authentication paths.

Credential Access

High

• Category: Privilege Escalation

Content

if grep -qiE "secret|password|token|key|credential" "$file" 2>/dev/null; then
      echo "  SENSITIVE: $file (contains credential patterns)"
    fi
  done | tee "$AUDIT_DIR/plaintext-secrets.txt"

# Check log files for PII leakage
echo ""

• Confidence: 92% confidence
• Finding: The skill scans .env, credentials, and secrets files for credential-related patterns and records sensitive file locations. This is dangerous because it directly targets likely secret stores and produces an inventory of where plaintext credentials may exist, which is highly sensitive if mishandled.

Credential Access

High

• Category: Privilege Escalation

Content

# World-readable sensitive files
echo ""
echo "--- World-Readable Sensitive Files ---"
find /home -type f \( -name "*.env" -o -name "*.pem" -o -name "id_*" -o -name "*.key" \) -perm /o+r 2>/dev/null | \
  while IFS= read -r file; do
    echo "  ❌ WORLD-READABLE: $file"
  done | tee "$AUDIT_DIR/world-readable-sensitive.txt"

• Confidence: 93% confidence
• Finding: The world-readable file scan targets .env, .pem, id_*, and .key files under /home, which are classic credential and private-key locations. Even though the goal is defensive, enumerating and recording these paths materially increases exposure of sensitive authentication assets.

Static analysis

No suspicious patterns detected.