evo-subagent
Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution — prefercmd/memory as knowledge allow/deny lists, so repeated tasks start from proven commands and save tokens
24 results
Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution — prefercmd/memory as knowledge allow/deny lists, so repeated tasks start from proven commands and save tokens
Self-evolving memory + skill lifecycle for DeepSeek Harness. Cross-session memory with zero-token deterministic recall (bigram-Jaccard fused with FTS5 BM25 via RRF), a tiered approval gate, and reinforcement that strengthens what you repeat. Procedural knowledge crystallizes into SKILL.md files that refine in place and are curated through an active-stale-archived lifecycle (reversible archive, pre-op backups, rollback, never deletes). Includes background per-turn review, anti-bloat convergence for both skills and memory, an auto-grown user profile, and a web settings page.
DSH web GUI notification alerts: distinct synthesized tones or a spoken voice (zh/en) for 'needs approval', 'needs answer', 'output complete' and 'error', with per-type sound, enable, volume, repeat, browser notifications and an i18n interface-language setting.
Cross-session agent-to-agent messaging for DeepSeek Harness: claims, verification and a decision ledger, so two agent sessions don't repeat, contradict or deadlock each other.
Research-debt guard for DeepSeek Harness: nudges coding agents to search docs, GitHub and the web before repeated local trial-and-error.
Session execution analytics and audit reports for DeepSeek Harness. Understand how your agent actually worked: steps, tool calls, failures, repeated actions, token usage and verification signals.
Two-level model circuit breaker with failover for DeepSeek Harness: trip a model or a whole provider after repeated request failures and route the next request to a configured fallback
On-time and valley-hour (off-peak) scheduled tasks with daily repeat for the DeepSeek Harness Web profile.
Hard-stop consecutive identical DeepSeek Harness tool calls after a configurable streak.
DeepSeek Harness task-protection plugin: plugin preflight scanner (catches missing `inject` declarations that crash the host), loop detection (repeated identical tool calls -> steer to a new approach), recursion guard (nested subagent depth cap), and interruption awareness (status route with safeToRestart). Never terminates tasks — only steers via injected messages. Zero npm dependencies.
Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution — prefercmd/memory as knowledge allow/deny lists, so repeated tasks start from proven commands and save tokens
AI turns repeated dev commands, prompts, and instructions into one-click DeepSeek Harness actions.
DeepJIT: JIT-compile repeated agent workflows into skills and flow templates for deepseek-harness (dsh)
Stop runaway agent loops in DeepSeek Harness: denies repeated identical tool calls and enforces a per-agent call cap, outside the model where instructions cannot reach.
Real-time repetition guard for DeepSeek Harness (DSH): stops model generation when the same string repeats >=10 times in the streamed output. 实时检测 DSH 大模型流式输出中的重复内容,同一字符串重复十次以上立即停止生成。
Configurable user-confirmed lessons that help DeepSeek Harness avoid repeating mistakes.
DeepSeek Harness plugin to heal third-party gateways that omit the stream terminal event (message_stop/done), which otherwise triggers repeated TRANSPORT retries in the dsh agent loop; scope is configurable per provider and per model.
重复字符串
Performance diagnosis, repeated plugin isolation campaigns, safe recovery, and measured verification for DeepSeek Harness
DeepSeek Harness plugin that blocks repeated identical failed tool calls
Result-aware stuck-loop guard for DeepSeek Harness: advisory nudges plus a monotonic hard stop. Only repeats with identical results count, so progressing polls and post-edit re-reads never trigger.
Strict gate: turns strict_check and the failure journal from tools the model may call into policy the harness enforces — repeated-failure intervention, a glob-protected critical-path gate, and an automatic post-write check.
Hard break on repeated identical tool calls: a synchronous ctx.tools.guard monotonic gate that denies the 2nd identical (tool name + canonical arguments) call per agent, with an optional same-path read guard and advisory first-repeat warning.
DSH (DeepSeek Harness) web plugin: adds a “PPT 制作” entry to the composer “+” menu that kicks off a bundled, self-contained PPT orchestration prompt — one menu pick, one short message, no repeated approval prompts.