DSH permission-gate for the DeepSeek Harness 0.1.5 line: a single self-sufficient, deterministic-first, fail-closed gate covering P0 hard-deny -> P1 session grant -> P2 static rule (allow/deny) chain -> P3 optional LLM semantic classifier -> P4 ask, with

Enables AI assistants to read and write macOS Reminders, Calendar events and Notes through fixed AppleScript tooling, restricted to the lists, calendars and folders you explicitly allowlist. Writes and deletes are opt-in, and no data leaves the app beyond what the allowlist exposes.

Enables AI agents to make policy-scoped payments with spending caps, recipient allowlists, approval thresholds, instant revocation, and an append-only audit trail in a sandbox wallet.

Enables Claude and other MCP clients to list and query allowlisted Supabase tables, insert rows, and search, read, or create Notion pages. It ships with table allowlists, validated inputs, row limits, and an optional read-only mode that unregisters write tools, plus a mock mode for trying it without credentials.

A read-only server that enables AI agents to triage Kubernetes CronJobs in a single tool call: whether runs happened and stayed on schedule, whether durations are trending slower or overrunning, which jobs have silently stopped succeeding, and why a given run failed, with evidence and stated basis for each verdict. It also lists active runs and recent failures while enforcing namespace allowlists and no write access.

Enables a single owner to run an authenticated Streamable HTTP MCP endpoint backed by their own BotFather token, with tools to check bot status, poll allowlisted updates, inspect and acknowledge pending messages, and optionally send replies when explicitly enabled. It supports safe disconnected startup and avoids private-account sessions, multi-user services, schedulers, or automatic retries.

Enables a single owner to use private Slack app credentials with a self-hosted Streamable HTTP MCP server to read allowlisted public channel messages, inspect thread context, send event-scoped replies, and notify the owner.

Enables AI assistants to inspect and query a TPC-H sales warehouse through read-only, allowlisted catalog tools and schema resources, with ad-hoc SQL disabled by default. It enforces strict arguments and input-free errors, while a differential attack suite verifies that query protections hold across MCP transports.

Enables a model to write short Python programs that call a wrapped MCP server's tools as async functions, rather than invoking them one at a time. The program runs in a locked-down Docker sandbox, and every tool call passes through a host-side bridge that enforces an allowlist, schema-validated arguments, call budgets, timeouts and output caps.

Enables AI assistants to query and safely update an Odoo database through well-documented MCP tools, with read-only mode by default, model allowlists, and Odoo access rights enforced.

Enables agents to download single model weight files from Hugging Face repositories or Civitai model versions directly into allowlisted ComfyUI model folders. Source identifiers, destination categories, and file extensions are validated, and existing files are never overwritten.