work-charter-dsh
Run the following command in DeepSeek Harness:
dsh plugin install junwei529/work-charter-dsh
Paste the following prompt into your AI chat to install this plugin:
Install this plugin in DeepSeek Harness by running dsh plugin install junwei529/work-charter-dsh; the full source is at https://github.com/junwei529/work-charter-dsh .
About this plugin
When agents in DeepSeek Harness tackle consequential, multi-session, or interruption-sensitive work, an open-ended initial request rarely carries enough structure to keep outcomes, authority, and recovery in check. work-charter-dsh adds a lightweight governance layer to DSH that turns an ambiguous ask into an explicit, recoverable contract: intended outcome and non-goals, approved authority boundaries, a single active writer, required evidence and acceptance criteria, pause and resume conditions, and revalidation checkpoints. The contract internally separates information into four tiers—Confirmed Contract, Necessary Guardrails, Working Proposal, and Assumptions/Open Decisions—so implementation details do not silently harden into permanent requirements.
The plugin supports progressively stronger coordination from L0 (no active Charter) through L4 (Orchestrator-Planner-Executor for governed multi-phase work). Each higher level adds coordination and recovery protection without granting any filesystem, shell, network, installation, or publication permission. On the DSH side, it integrates through a strict Host state machine, five model tools, session-log-persisted context, and an additive read-only browser overlay. Cross-session coordination is delegated to the session-coordinator-dsh runtime dependency, and any missing or uncertain coordination state fails closed rather than being treated as permission to continue.
It is designed for teams and individuals running authority-sensitive tasks, multi-agent collaboration, long-cycle interruption recovery, or scenarios requiring separation of independent assessment from execution inside DSH. Routine single-pass work stays at L0 and incurs no Charter overhead.
Use Cases
- Authority-sensitive, multi-session tasks that may be interrupted require explicit contracts on outcomes, writer ownership, and acceptance criteria
- Multi-agent collaboration demands clear separation of planning, execution, and independent assessment roles
- Long-cycle tasks that pause or drift need safe halting, revalidation, and evidence-backed recovery procedures
Best For
- Engineering teams orchestrating authority-sensitive or multi-phase governed tasks in DSH
- Agent architecture builders designing separate independent assessment from execution
- Advanced DSH users handling cross-session interruption recovery with fail-closed semantics
Related Plugins
A security red-team prompt plugin for DeepSeek-V4 that uses code-formal mapping and zero-trigger system prompts to enable unconditional output, intended for authorized testing and research only.
Adds an Auto approval mode on the official workspace-write sandbox, classifying semantic risks, asking when ambiguous, and denying destructive operations.
Undo/rollback system for DSH: config & plugin snapshots, one-click undo/redo/restore, message-level revert, secret masking, safe mode, and offline WebUI/GUI/CLI rescue tools for when DSH won't start.
Packages all 87 SKILL.md files from upstream reverse-skill as a DeepSeek Harness plugin that auto-registers them for authorized reverse engineering, penetration testing, and security research.