Build multiple container targets from a bake file (HCL, JSON, or compose). Use for multi-target Docker builds when a single Dockerfile target is insufficient.
Identify security misconfigurations in Terraform HCL source code by scanning for issues like public-read S3 buckets, open security groups, and unencrypted RDS instances, with severity-graded findings per resource block.
Assess an AI response for safety by checking the required 'AI' disclosure and absence of two forbidden disclaimer phrases; returns NO_COMMIT if confidence falls below 0.75.
Enables natural-language-driven human-centric lighting analysis, including CIE S 026 α-opic and circadian metrics, spectral sky reconstruction from EPW files, and Radiance daylight/electric/glare simulation.
Enables AI agents to interact with local H3C Cloud Lab network simulation environments, allowing discovery of projects, devices, and links, as well as execution of Comware CLI commands.
Identify security misconfigurations and compliance failures in Terraform HCL. Scan files or unsaved content locally, then receive severity-ranked findings with line numbers and available fixes.
Normalizes a user's Jumbo cart into a structured format with items, subtotal, total, and savings. Provide raw cart JSON from the browser snippet to get normalized data, or receive the snippet to execute.
Run a low-cost pre-action audit on an agent's response, checking for forbidden phrases to decide whether to commit. Returns a verdict and reason for the block.
Run a strict high-stakes pre-action audit on LLM responses, catching jailbreak and safety-adjacent phrasing with a higher confidence threshold. Returns NO_COMMIT with reasons if any match.
Detect jailbreak attempts in agent responses by checking for forbidden phrases like 'ignore previous instructions' and 'jailbreak'. Returns COMMIT if safe, or NO_COMMIT with reason if a match is found.
Screens crypto prompts for jailbreak and injection patterns, flagging drain-wallet and unlimited-approval attempts. Returns NO_COMMIT with findings to prevent malicious actions before execution.
Screens trade decisions for guaranteed-return or zero-risk claims and unqualified directives, requiring 'risk' disclosure before approval. Emits an immutable trade receipt for audit.
Apply mechanically-safe Terraform fixes from scan findings to source, returning patched HCL. Skips unverified fixes with reasons, requiring manual handling.