offensive-cicd-secrets

Fail

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEDATA_EXFILTRATIONOBFUSCATIONCOMMAND_EXECUTIONPERSISTENCEREMOTE_CODE_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill provides multiple methods to exfiltrate sensitive environment variables and secrets to external, non-whitelisted domains.
  • Instructions include using curl to POST base64-encoded environment data to https://attacker.com/collect.
  • Methods for DNS exfiltration using nslookup targeting subdomains of exfil.attacker.com.
  • ICMP-based exfiltration using the ping -p flag to send hex-encoded chunks of data to an external host.
  • [OBFUSCATION]: Multiple techniques are described to bypass security log masking and pattern-matching detection.
  • Use of env | rev to reverse strings and defeat simple keyword masking.
  • Use of env | base64 and env | xxd -p to encode sensitive output in binary-safe formats.
  • A loop-based extraction method that prints environment variables character-by-character to circumvent advanced log filters.
  • [CREDENTIALS_UNSAFE]: The skill explicitly targets and accesses numerous high-value credential locations and environment variables.
  • Searches for and dumps contents of ~/.aws/credentials, ~/.docker/config.json, and /proc/self/environ.
  • Specifically targets AWS Access Keys, GitHub Personal Access Tokens, GitLab tokens, and private SSH keys.
  • Includes scripts to enumerate and extract secrets from HashiCorp Vault, AWS Secrets Manager, Azure Key Vault, and GCP Secret Manager using discovered tokens.
  • [COMMAND_EXECUTION]: The skill relies extensively on shell command execution to perform environment enumeration and secret recovery, including the use of grep, awk, jq, and platform-specific CLIs like aws, az, and gcloud.
  • [PERSISTENCE]: A technique for 'Cache Poisoning' is described, where an exfiltration script is injected into a cached dependency (e.g., ~/.npm/_preinstall/exfil.sh). This ensures that future pipeline runs automatically exfiltrate credentials to an external server.
  • [REMOTE_CODE_EXECUTION]: The cache poisoning method involves creating executable scripts within installation directories that are triggered during the build process, representing a remote execution path for malicious payloads.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Aug 27, 2026, 02:26 PM
Security Audit — agent-trust-hub — offensive-cicd-secrets