DeepSeek Harness PluginManifest Valid

dsh-codex-approval

Codex-style approval autopilot for DeepSeek Harness combining rule matching and AI risk assessment.

Terminal & TUIDeveloper ToolsSecurity & Policy

2GitHub Stars0ForksUpdated2026-10-10

Source

Install

$ dsh plugin --profile web add dsh-codex-approval

Plugin 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.

source-grounded

dsh-codex-approval is an automated approval plugin for DeepSeek Harness designed to reduce interruption prompts during agent execution. It processes command execution requests using ordered glob rules alongside a tiered AI risk judge. Read-only commands are allowed automatically, destructive operations are prompted or denied, and ambiguous actions are assessed by an evaluation model with fallback and unattended execution support.

Key Capabilities

Ordered rule matching to instantly allow safe read-only commands or block destructive operations without invoking model calls
Tiered AI risk evaluation for ambiguous commands, with support for session context and primary/fallback judge models
Multiple operational approval modes: interactive AI mode (ai), unattended auto mode (ai-auto), and manual override (manual)
Built-in parsing and rule matching for both Bash and Windows PowerShell commands
Audit logging of all approval and rejection decisions with reasons recorded in JSONL format
Web settings panel integration for configuring judge models, fallback models, and default behaviors

Useful For

Running long-running or autonomous agent workflows without getting blocked on routine read-only command confirmations
Enforcing strict safety policies against catastrophic file deletions and unrecoverable git/docker/kubectl state wipes
Unattended background execution where risky or uncertain actions are automatically denied rather than hanging indefinitely
Auditing and reviewing automated agent command execution decisions in workspace environments

Who It Fits

DeepSeek Harness users seeking to automate repetitive permission prompts during agent sessions
Developers running long, multi-step tasks in workspace-write sandboxes
Security-conscious operators requiring explicit guardrails against irreversible workspace or repository destruction

Documented Limitations

Requires Node.js runtime version 22.19 or higher
Only intervenes in workspace-write or read-only modes; inactive under danger-full-access permissions where approvals are not generated
Depends on accessible local or remote judge models; defaults back to manual approval if all judge models fail
Running multiple approval plugins concurrently on the same profile can cause conflicts

DSH Compatibility

Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.

Not tested yetNo runtime compatibility tests have been published yet.

Security Signals

Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.

Dependency Counts

package.json declares 1 runtime, 0 development, 0 peer, and 0 optional dependencies.

info

Dsh Bundle Declared

package.json declares DSH bundle metadata.

info

License Declared

GitHub reports the repository license as MIT.

info

Package Manifest Available

A root package.json was captured and can be inspected by the registry.

info

Source Available

Public GitHub source metadata is available for this registry snapshot.

info

Source & Registry Notes

Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.

Source repository: 040822/dsh-codex-approval
Registry source: GitHub · dsh-plugin topic
Registry classification: Plugin
Source checked: e64b07a · 2026-10-24

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.