dsh-auto-review
Second-model AI auto-review for DeepSeek Harness approval requests with fail-closed security.
Install
$ dsh plugin --profile <name> add dsh-auto-reviewPlugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Intercepts approval/request waterfall events to selectively auto-review sensitive tool operations
- Spawns an isolated read-only reviewer subagent with restricted tools (read, glob, grep)
- Fails closed on timeouts, crashes, or schema mismatches using configurable fallback policies
- Injects auditable rejection reasons directly into denied tool results
- Supports session-level circuit breaking after consecutive denials
- Provides session projection and web review panel UI for runtime monitoring
- Includes dsh-eval test engine and a standalone stdio MCP server
Useful For
- Automating tool execution approval (e.g. bash or file writes) without constant manual human prompts
- Preventing dangerous terminal commands using regex risk rules and second-model analysis
- Auditing agent decision traces through structured session-log events
- Regression testing agent trajectories and prompts via the included dsh-eval platform
Who It Fits
- DeepSeek Harness developers running autonomous agent workflows
- Security-focused teams operating sandboxed AI developer agents
- DevOps and CI/CD engineers automating tool execution with strict guardrails
Documented Limitations
- The reviewer model is an advisory evaluator rather than a cryptographic security kernel
- Requires a functional LLM route configured for the reviewer subagent
- Tools listed under reviewerTools must exist as global tools in the active profile
- The standalone MCP server path does not invoke a second model and relies only on deterministic rules and cache
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 2 runtime, 46 development, 15 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as Apache-2.0.
A root package.json was captured and can be inspected by the registry.
package.json declares a prepare lifecycle script that may run during relevant package installation workflows.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- PerryLink/dsh-auto-review
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- eda595c · 2026-10-23
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 234GitHub stars
- Forks
- 3
- Open issues
- 2
- Last commit
- 2026-10-22
- Last release
- 2026-10-22
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Native SQLite-backed task management and project board bundle with Web UI and Agent workflow tools for DeepSeek Harness.
View plugin →Persistent multi-model agent teams with dynamic DAG orchestration, run monitoring, and token insights for DeepSeek Harness.
View plugin →Agent skill for DeepSeek Harness that turns ideas, plans, and codebases into interactive architecture diagrams.
View plugin →Agent-native multi-market trading terminal and quantitative research plugin bundle for DeepSeek Harness.
View plugin →