agent-infrastructure

Privacy-minimal heuristic per-turn verification summaries for DeepSeek Harness

Observe-only hash-chained evidence receipts for DeepSeek Harness

Declarative deny-by-default tool policy plugin for DeepSeek Harness

Shared lifecycle admission protocol and reference policy kernel for DeepSeek Harness subagents.

DSH-Plugin: structural observability layer - real tool/skill usage recording (tools/result), skill-catalog audit, watchdog. No model self-reporting.

DSH-Plugin: durable agent memory core - mem_save/mem_search/mem_health backed by storageDomain (survives restarts), 記憶 view tab.

Portable operations skill for diagnosing and safely remediating DeepSeek Harness runtime issues.

DSH-Plugin: multi-engine cross-validated search - parallel Brave/Tavily/DDG/Wikipedia, URL dedupe, confidence tiers.

DSH-Plugin: minimal infrastructure entry - infra_status/infra_bootstrap probes for the mounted agent-infrastructure stack.

Diagnose poisoned DeepSeek Harness sessions and continue from provider-safe recovery boundaries

DSH-Plugin: skill trigger/selection gate - user picks from described candidates when the model triggers a skill; 10s no-choice falls back to the system default.

DSH-Plugin: per-reply skill trail - truthful per-turn triggered/used tool & skill badges, folded from the session log (no model self-reporting)