A DeepSeek Harness plugin that stops agents from drifting off-task: it locks in the good first-round behavior, watches every step, and pulls the agent back when it degrades.
让 Agent 查询课程能力、清单知识点和本地学习者档案,辅助判断下一步学习。 | Let agents query curriculum capabilities, checklist knowledge, and local learner profiles to guide next steps.
Solidify DSH session history step by step: rewrite tool results in place to the facts the model actually kept, so long conversations stop re-sending noise every turn. Costs tokens and time: a ~770-character contract in every request, plus one extra full-context request and round-trip per step when stepSummary is on.
DeepSeek Harness plugin: when a run finishes or needs approval, get a link on your phone, hear the result from an ElevenLabs voice agent and say what to do next
Auditable mathematical proving for DeepSeek Harness: a local stdio MCP server (23 tools), a bundled workflow skill, and the /prove and /audit-proof commands. proven / refuted / inconclusive are never conflated, and unproven steps are reported, not hidden.
Run an agent unattended for hours and get a morning brief that is an audit, not a summary. Signs a manifest before the work, prices every step at the execution boundary, rolls back a failed verification, and reports what the gate refused. DSH plugin + skill, zero dependencies.
DSH plugin: hands a long session over to a fresh session in the same workspace on /handoff or at a token threshold, carrying the preset, model, permissions and plan mode across; the brief is written by the session's own model and the switch waits for a step boundary instead of interrupting running work.