Content
76%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
An exceptionally lean, command-dense playbook with a clear numbered progression from cluster identification through privilege escalation to host compromise and reporting. The main gaps are the absence of validation checkpoints before risky batch/destructive steps, a literal-ellipsis placeholder in the permission-inventory command, and undefined internal pseudo-API calls (k8s_audit, kg_add_node).
Suggestions
Add explicit validation checkpoints before destructive/batch steps: e.g. confirm the escape pod is Running before exec ('kubectl get pod escape-pod --watch'), and test a stolen token with a single low-impact API call before batch-harvesting all secrets.
Replace the 'cat .../namespace' ellipsis placeholder with the actual path '/var/run/secrets/kubernetes.io/serviceaccount/namespace' so the inventory snippet is copy-paste executable.
Briefly define the reporting interface (k8s_audit, kg_add_node, kg_add_edge) or point to a reference file so the 'Decepticon ingest' and 'Promote' steps are actionable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean and command-dense throughout: every section is executable snippets or tight reference tables with zero concept explanations ('what Kubernetes is', 'how RBAC works') and no padding — every token earns its place. | 5 / 5 |
Actionability | Nearly all guidance is copy-paste executable (kubectl invocations, a complete escape-pod YAML, jq decoding pipelines), but 'cat .../namespace' uses a literal ellipsis placeholder instead of the real path, and the 'k8s_audit()' / 'kg_add_node()' pseudo-API calls are referenced without any definition or context, leaving minor gaps. | 4 / 5 |
Workflow Clarity | Sections 1-8 form a clear logical sequence (identify cluster -> inventory permissions -> RBAC escalation -> pod escape -> secrets harvest -> api-server pivot -> promote), but this skill involves batch and destructive operations (batch secret exfiltration, privileged escape-pod creation) with no explicit validation or verification checkpoints, capping workflow clarity at 3 per the judging guidelines. | 3 / 5 |
Progressive Disclosure | A single-file skill (no references/, scripts/, or assets/ exist) organized into well-labeled numbered sections with no buried or nested references. Minor gap: the CVSS and defender-remediation sections are tangential to the operator workflow and could live in a separate reference file, though at ~165 lines total the inline placement is defensible. | 4 / 5 |
Total | 16 / 20 Passed |