Use DeepSeek Harness's built-in tools and capabilities before installing third-party plugins or inventing shell/custom workarounds. Use whenever a task may be handled by native DSH files, search, shell, jobs, LSP, subagents, workflow, Ralph, goals, todo, web, sessions, planning, or other shipped capabilities; also use…
A recovery guide for common DeepSeek Harness failures, such as missing commands, stale text, unavailable history, or failed installation. It examines the current error and applies a small, matching recovery step.
Add, review, or update a bansos.dev listing using the current folder-based catalog and contributor profile flow. Use when an AI agent needs to verify a program source, run npm run add:bansos, create or update a listing folder, attribute a first-time or existing contributor, validate data, or prepare and review a…
Use when an agent needs live, structured search-engine data through SerpApi (web results, news, shopping, local, jobs, flights, scholar, images, video). Covers picking the right engine for the intent, choosing compact vs complete output, query operators, pagination, deduping and citing results, and when NOT to search.…
Skill "intern-os" from poktalabs/intern-os, covering internos — workstreams (claude code), what "thread" means in claude code, resolution: pwd → workstream, returns workstream absolute path on stdout (exit 0), or and operating protocol.
Run a coding agent unattended overnight, long running agent safety, stop a runaway loop, cap agent token spend. Sets up four guardrails around an agent loop you already have: a resumable handoff file so an interrupted run continues, a verification gate that fails closed so unchecked work is never accepted, a runaway…
Build, inspect, edit, schedule, price, place, and cancel Amazon Fresh grocery orders through the local fresh CLI. Use when a user asks an agent to search Amazon Fresh, choose groceries, create or modify a cart, compare prices, select an address or delivery slot, configure substitutions or tip, preview checkout, place…
Use when diagnosing Claude Code prompt-cache breaks, 1h cache misses, cachecreation/cacheread anomalies, away summary or recap rebuilds, cc-switch proxy logs, or reports that Claude Code recreates cache after a few minutes.
Use when creating, adding, replacing, or polishing desk pet roles for the Claude Code desk pet project, especially when the user describes a role and expects the agent to generate prompts and guide the whole workflow.
A controlled workflow for manually synchronising one Markdown file with its linked Feishu document, a collaborative document in Lark’s productivity suite.
A Chinese writing-style guide for producing very short, fragmented passages, with one sentence per paragraph and emotion conveyed through objects, numbers, and silence rather than stated directly.
A live role-playing experience set in an open world called 留风镇. Each turn describes the surroundings, the character's thoughts, choices, and the results of actions, with a progress line included.
Reference and implementation guide for building Discord bot UIs with Discord.Net's Message Components V2 system (ComponentBuilderV2) in C#. Covers every V2-only component (Container, Section, Text Display, Thumbnail, Media Gallery, File, Separator), the shared V1/V2 interactive components (Action Row, Button…
Manage ML/LLM experiment workflows with ztxexp. Use when the agent needs to build parameter configurations, run experiments, validate run artifacts, analyze results, or troubleshoot failed runs.
★not rated 4▲
+1 6mo agoA42 tokens
originalMIT
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: