The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill requires elevated privileges for hardware interaction and diagnostic procedures which is typical for this domain.
Evidence: TASKS.md and CAPABILITIES.md specify that DPU-side process privileges (typically sudo) are required for PCIe-level emulation and for executing dmesg to inspect host-side kernel logs.
[INDIRECT_PROMPT_INJECTION]: The workflow creates an indirect prompt injection surface as the agent is instructed to ingest and process outputs from external system commands.
Ingestion points: Diagnostic steps in TASKS.md and CAPABILITIES.md rely on the agent reading outputs from lspci, dmesg, and doca_caps.
Boundary markers: The instructions lack explicit guidance on using delimiters or ignoring instructions embedded within these external outputs.
Capability inventory: The agent is guided to execute shell commands and modify build manifests based on the ingested data.
Sanitization: No specific sanitization or validation of the command outputs is prescribed.
[DYNAMIC_EXECUTION]: The skill guides the user through building and running compiled binaries which is the primary intent for library development.
Evidence: TASKS.md defines workflows for compiling C/C++ source code using pkg-config and executing the resulting DPU-side binaries for testing and operation.

doca-devemu