The Agent Skills Directory

[SAFE]: The skill provides instructional guidance for analyzing system capacity, workload modeling, and resilience planning without any malicious patterns detected.
[SAFE]: Instructions explicitly forbid the inclusion of secrets, customer data, tokens, or raw sensitive traces in the output, demonstrating good security hygiene.
[SAFE]: The skill includes strong warnings against unauthorized load testing, resource creation, or quota modifications in production environments.
[SAFE]: External references target official documentation from well-known technology organizations (Kubernetes and Grafana) for educational purposes, posing no security risk.
[INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to process external repository content like runbooks, ADRs, and incident reviews. This represents an attack surface where untrusted data could attempt to influence the agent's reasoning. However, this risk is mitigated by explicit instructions to categorize inputs (measured vs. assumed) and strict non-negotiable rules against destructive actions or secret exposure. Ingestion points: repository instructions, runbooks, and incident reviews. Boundary markers: explicit categorization of inputs as measured, reported, derived, or assumed. Capability inventory: repository and production system inspection (analysis-focused; no automated resource modification tools allowed). Sanitization: prohibitions on including secrets or customer data in output.

headroom