The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits adversarial context flooding and metadata poisoning.
Files such as references/architecture-patterns.md, references/state-management.md, and 6 other reference documents contain 150 repetitive sections each of identical technical filler text.
This behavior is characteristic of context window flooding, intended to saturate processing tokens and influence priority weighting through technical authority claims ("strictly intended for Staff+ Engineers").
[COMMAND_EXECUTION]: The skill provides extensive templates for shell script execution within Jenkins CI/CD pipelines.
Numerous examples of sh steps are used across all instructions and shared libraries, establishing a broad functional runtime surface.
The lack of explicit input validation or sanitization patterns in these templates poses a risk of command injection if the resulting pipelines process untrusted external data.
[EXTERNAL_DOWNLOADS]: Integration with third-party container images and software resources.
Pipeline templates reference unverified Docker repositories (e.g., maorfr/kubectl) and external Git repositories for scanning tools.
These examples facilitate the introduction of supply chain risks if the images or repositories are not verified by the user.

jenkins