planning-disaster-recovery
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied infrastructure inventory and service requirements to generate recovery plans and automation logic. This introduces a surface where adversarial data in the input could attempt to influence the logic of the generated output.\n
- Ingestion points: User input defining production services, dependencies, and criticality tiers as described in
SKILL.md.\n - Boundary markers: The instructions in
SKILL.mddo not implement explicit delimiters or warnings to treat inventory data as untrusted.\n - Capability inventory: The skill is granted access to file writing and editing tools, along with restricted bash commands for
tar,rsync, andaws s3operations inSKILL.md. The scripts described inscripts/README.md(dr_plan_generator.pyandsetup_code_generator.py) likely perform the primary file-writing and code-assembly tasks.\n - Sanitization: The skill does not specify sanitization or validation steps for user-provided infrastructure strings before they are interpolated into the generated scripts.\n- [DYNAMIC_EXECUTION]: The skill generates infrastructure-as-code (Terraform) and automated failover scripts at runtime based on dynamic user requirements.\n
- Evidence: The
SKILL.mdoverview and instructions 4 and 7 explicitly require the generation of Terraform modules and automated failover scripts for databases, DNS, and compute scaling.
Audit Metadata