dsh-action-outbox
Run the following command in DeepSeek Harness:
dsh plugin install JimChen-g/dsh-action-outbox
Paste the following prompt into your AI chat to install this plugin:
To install this plugin in DeepSeek Harness, run `dsh plugin install JimChen-g/dsh-action-outbox` or visit the source code at https://github.com/JimChen-g/dsh-action-outbox.
About this plugin
Agents often prioritize speed by calling external APIs directly, leading to irreversible actions like sent emails or deployed code. This irreversibility poses significant risks in critical workflows. dsh-action-outbox solves this by introducing a "stage, review, commit" workflow that separates execution from side effects.
The plugin's core strength lies in its ability to decouple external writes from execution. Agents can stage tool calls, but they must pass through a rigorous review process before committing. It features a batch review inbox that allows inspecting complete canonical JSON, byte counts, and tool fingerprints, ensuring every action is verified. With SHA-256 digests and single-use approval nonces, it provides robust safety semantics to prevent race conditions and tampering.
This tool is ideal for DSH developers who require strict audit trails and control over unintended side effects, particularly in scenarios involving financial transactions or sensitive data. It transforms external operations from "fire-and-forget" to a verifiable and reversible process, granting human oversight without the overhead of distributed transaction management.
Screenshots
Use Cases
- Prevent irreversible side effects like deployments or emails from agents.
- Manually review and edit critical operations.
- Keep complete logs and parameter snapshots of tool calls.
Best For
- DSH workflow developers needing a 'fail-closed' strategy.
- Teams handling sensitive data writes or payment logic.
- Enterprise users seeking high security and auditability.
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.