The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or persistence mechanisms were detected. The skill is security-centric and provides authoritative guidance on hardening DevOps infrastructure.
[EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading and installing CI/CD runner software from official, well-known registries.
Evidence: references/runners.md includes installation scripts for gitlab-runner (via packages.gitlab.com) and actions-runner (via github.com/actions). These sources are recognized as trusted/well-known technology services.
[COMMAND_EXECUTION]: The skill manages the generation and auditing of pipeline configurations that execute shell commands on build runners.
Evidence: Examples in references/github-actions.md and references/gitlab-ci.md demonstrate the use of run: and script: blocks. The skill enforces security posture by requiring SHA pinning and environment variable assignment to prevent command injection from external events.
[INDIRECT_PROMPT_INJECTION]: The skill identifies and provides mitigations for indirect prompt injection surfaces where pipelines process untrusted user data.
Ingestion points: github.event.* fields (e.g., issue titles, PR descriptions) as documented in SKILL.md and references/github-actions.md.
Boundary markers: The skill explicitly instructs agents to assign untrusted data to environment variables before shell use to prevent macro-expansion/injection.
Capability inventory: Shell execution capabilities across all target platforms (GitHub, GitLab, Forgejo, etc.).
Sanitization: The skill provides a specific 'AI Self-Check' section to ensure generated pipelines do not interpolate untrusted data directly into command blocks.

ci-cd