AI Agent runtime authorization & evidence verification — tool-call GuardrailProvider, CCS 7-dimension verification standard, MCP/DSH security scanner, SSRF/command-injection/credential-exfil blocking with Ed25519 signed receipts.

Git-native slide format, Agent API, CLI, MCP server, and renderer

Three-tier memory control plane for DeepSeek Harness: persistent runtime context, searchable project documents, pluggable long-term memory, smart routing, supervised agent workflows, WebUI, and headless tools.

DeepSeek Harness session cost meter plugin: session/daily cost, budget, history, OpenCode Go quota, official & custom-provider balance, Codex-like token heatmap, peak/off-peak pricing with pre-switch popup & system-notification alerts, official price sync, 90+ model pricing catalog, Coding Plan quota queries (7 vendors), bilingual zh/en UI

Model-driven context management (Active Context Pruning / ACP) for the DeepSeek Harness — the model decides when and what to compress. Ported from billion-context-pi (ranxianglei); acp-kernel reused verbatim. CompactionEngine backend with compress/decompress/search_context/acp_status tools.

Second-model AI auto-review for DeepSeek Harness approval requests: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default, fully auditable from the session log (approval/asked -> autoReview/verdict -> approval/decided).

Bounded, layered, approval-gated, auditable cross-session memory for DeepSeek Harness (capability seam: ctx.memory + SQLite provider + memory tool + frozen snapshot injection)

Auditable vision and cross-platform Computer Use runtime for DeepSeek Harness with source-preserving evidence.

Open-source self-improving QA agent for software teams. A test harness with memory. Write tests in natural language for web and mobile. agent-qa learns from every run, adapts to UI changes, and catches regressions before you ship.

Delete DeepSeek Harness sessions from the UI: header danger button + sidebar session-row menu item (no conversation jump), risk-consent dialog with session name/id, stops running agents first, in-place list refresh without page reload. Works in web and the desktop client.

Runtime evidence that helps agents trace, profile, and burn down hotspots in application and native code, GPU kernels, and inference stacks.