dd-apm
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the Datadog Agent installation script from the official
install.datadoghq.comdomain (e.g., inlinux-ssi/agent-install/SKILL.md). - [EXTERNAL_DOWNLOADS]: Downloads the
pupCLI tool directly from the author's (datadog-labs) GitHub repository and official releases. - [COMMAND_EXECUTION]: Executes system-level commands using
helm,kubectl, andsshto install, configure, and troubleshoot the Datadog Agent and application instrumentation. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes environment metadata (such as service names, pod labels, and namespaces) which is subsequently used in shell commands for discovery and verification.
- Ingestion points: Reads output from
kubectlandpupdiscovery commands (e.g.,SKILL.md,troubleshoot-ssi/SKILL.md). - Boundary markers: No explicit delimiters are used when interpolating discovered variables into prompts or shell commands.
- Capability inventory: Utilizes high-privilege tools including
ssh,kubectl,helm, anddockeracross the workflow. - Sanitization: Discovered infrastructure identifiers (like service names) are used directly without explicit escaping or validation before being passed to subsequent tool invocations.
Audit Metadata