dsh-project-anchor
Run the following command in DeepSeek Harness:
dsh plugin install KKLL2025/dsh-project-anchor
Paste the following prompt into your AI chat to install this plugin:
In DeepSeek Harness, run dsh plugin install KKLL2025/dsh-project-anchor to install this plugin; the full source is available at https://github.com/KKLL2025/dsh-project-anchor .
About this plugin
A long, multi-session DSH project is hardest at cold start: a fresh session rarely knows which Goal is still active, which Milestone is being validated, or where the last Turn checkpoint stopped. dsh-project-anchor pins the project to a user-approved Goal, one active Milestone, and a durable Turn checkpoint, so each recovery starts from verified state rather than model inference.
The plugin delegates persistent state to DSH's native layer and project judgment to on-demand Skills. It remembers only consent, the stable Goal association, and a checkpoint tool exposed while the Anchor is active; it does not infer or rewrite the user's decisions. When drive-large-project loads, it selects the lightest sufficient execution intensity—fast, continuity, or high-risk/release—based on current risk and acceptance boundaries. That intensity remains temporary Skill judgment instead of a persisted plugin field, so the model cannot raise it through an Anchor state field.
It suits solo developers or small teams running multi-day projects in DSH without adding a separate planner or risk field. Cold recovery is consent-gated: only an explicit Resume answer triggers DSH's native resume, while deferral, missing state, or drift leaves the Goal, Anchor, and Milestone unchanged. The plugin is distributed as a pinned GitHub source artifact rather than an npm package, which is appropriate for users willing to review the fixed commit SHA recorded in each preview release.
Use Cases
- Drive a multi-session project with automatic Goal and Milestone recovery on every cold start
- Cross-module or cross-Turn work where checkpoints ensure resumability after interruption
- High-risk release scenarios where execution intensity escalates with a full evidence trail
Best For
- Solo developers driving multi-day, multi-session projects in DSH
- Small teams preferring lightweight persistence over a separate planner
- Users who want consent-gated recovery and guard against silent state changes
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.