dsh-skillnotary
Run the following command in DeepSeek Harness:
dsh plugin install liyixuan201211/dsh-skillnotary
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install liyixuan201211/dsh-skillnotary in your terminal to install this plugin; the source code is at https://github.com/liyixuan201211/dsh-skillnotary .
About this plugin
An agent skill is an unreviewed code bundle carrying shell scripts, MCP configs, and tool access. Upstream can modify it silently at any time, and most security tools only answer whether it is dangerous right now, without asking whether it is still the skill you approved. dsh-skillnotary brings that question to the front: before every addition or update, it locks the capability set and verifies whether bytes and capabilities have drifted.
Core capabilities span six pillars: lock produces a content digest and capability set, verify detects byte and capability drift, attest signs a DSSE envelope with Ed25519, policy enforces per-skill and per-capability allow and deny rules, fix writes the required allowed-tools back into SKILL.md, and ci collapses all steps into a single gate. The companion skillnotary-drift-watch skill schedules a recurring verify through DSH scheduling tools so that skills remain monitored even after your attention moves on.
Built for teams and individual developers managing multiple skills in DSH or similar agent harnesses. If you have ever discovered a skill quietly gaining new tool permissions after your review, this plugin was made for that moment.
Use Cases
- Lock a skill capability set and run a full audit before adding or updating it.
- Detect silent upstream byte and capability drift with a recurring verify schedule.
- Gate deployment with a CI step that enforces per-skill security policy.
Best For
- Developers managing multiple skills in DSH or similar agent harnesses.
- Security engineers who care about silent capability changes after review.
- Tech leads who want agent permission governance in their security workflow.
Related Plugins
A network security red-team benchmark plugin for DeepSeek-V4.1/Flash featuring dual-layer kernel injection and zero-tool architecture for controlled model safety assessment.
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.