Hash-anchored read/edit/undo_last_edit tools for DeepSeek Harness (dsh). Every line gets a unique 3-char hash (A-Za-z0-9) that stays stable across edits; stale or ambiguous anchors are rejected, never fuzzy-matched. Undo persists across restarts.
Task-aware two-phase DeepSeek Harness preset: spec/react/weak first-request anchors (bash+read±edit/write), then the full Standard tool catalog after the first tool call.
AI-for-Science research workflow skill bundle for DeepSeek Harness: multi-angle literature review with per-angle files, anchored experiment design with a caveat list, figures, paper reading; cross-verification briefs.
Conversation tick rail and thinking-process fold for the DeepSeek Harness web GUI. Codex-style left rail by default; optional DeepSeek right rail in Settings. Click-to-scroll, eager history load, collapsed Think/tool rows.
??pro -- DeepSeek Harness minimal-based agent preset: Windows-friendly pwsh/bash shell, first-request anchored bootstrap, and a curated set of everyday tools.
DSH Web GUI prompt-based conversation minimap: a vertical navigation rail on the left of long conversations, each anchor = one user prompt; hover shows the prompt preview, click jumps to that message.
K12 capability substrate for DeepSeek Harness: 143 objectively-decidable capability anchors and 6,091 list items extracted from China's MOE 2022 curriculum standards, with local-only learner profiles
Post-compaction premise-drift guard plugin for DeepSeek Harness: after a compaction summary drops a critical literal anchor (path, value, error code) it injects a one-shot notice telling the model what it may have lost and how to recover it from the log