The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses neuron-explorer to start a local database server and curl to perform queries. These actions are legitimate and confined to the local environment. \n- [DATA_EXFILTRATION]: All network requests are made to localhost. There is no evidence of data being transmitted to external or untrusted domains. \n- [EXTERNAL_DOWNLOADS]: The skill references standard AWS Neuron software packages (aws-neuronx-tools), which are appropriate for the target hardware environment. \n- [PROMPT_INJECTION]: The skill ingests performance trace files (NEFF, NTFF) as a core data analysis task. \n
Ingestion points: External files processed by the neuron-explorer view command in SKILL.md. \n
Boundary markers: The agent interacts with the data through structured SQL queries and Pandas DataFrames, providing inherent data boundaries. \n
Capability inventory: Capabilities include local shell command execution (via curl and neuron-explorer) and Python script execution for data processing. \n
Sanitization: Data is processed and normalized by official vendor utilities before being accessed by the agent. \n- [SAFE]: The skill demonstrates no malicious intent, obfuscation, or persistence mechanisms. Its behavior aligns with its stated purpose of hardware performance profiling.

neuron-nki-profile-querying