🔌
Plugins
Loading...
Showing 30 of 14294 results
from AojdevStudio/dev-utils-marketplace
from AojdevStudio/dev-utils-marketplace
from Angriff36/task-orchestrator
Makes Claude Code's native task list a live work queue with event-driven hooks and zero per-prompt cost. A SessionStart hook primes the model to treat the native list as the queue, re-surfaces a repo's unfinished tasks from earlier sessions (the native list starts empty each session), and hydrates from a committed roadmap/backlog when present. On any substantive prompt (multi-step, or consequential/irreversible) a UserPromptSubmit hook runs the interpret→present→approve loop: the model first EVALUATES the request instead of just executing it — steelmanning then challenging it, flagging any contradiction with the project's recorded constraints OR the owner's own earlier requests, and any constraint/bias forcing a poor or over-engineered design, and recommending against part or all when that is its honest read (selective — only on real signal, since manufactured objections train rubber-stamping); then it interprets the request, decomposes it, judges each task for risk/alignment and parallel-vs-inline fan-out, presents its understanding + candid per-task recommendations (including skip) via AskUserQuestion, scaled to the work (a brief inline confirmation for small low-risk work; the full AskUserQuestion present-and-approve for larger or consequential work), and TaskCreates only what the user approves — weighed against the project's recorded direction. Consequential prompts get extra scrutiny; trivial prompts stay silent and run under auto mode. On a visual/UI/layout prompt (precision-tuned heuristic — architecture/API 'design' and functional edits don't trip it) the loop specializes into a design preview: the model presents a recommended design plus 2-3 alternatives as faithful ASCII mockups in the AskUserQuestion preview (native arrow-key nav + Enter, recommended first) and builds only the chosen one — the owner-loop's demonstrate-before-build for a non-technical owner who verifies by seeing. Non-blocking (it only injects the instruction). A Stop hook CLOSES the loop with an intent→outcome gate: the substantive prompt's plain-language ask is stashed (the intent of record) and replayed at 'done' against the actual diff, blocking once (consumed per ask, so it can't loop) so the model verifies the OUTCOME matches the request — surfacing 'built the wrong thing / only part of it / something extra' in plain language to the non-technical owner (who verifies by seeing, not reading code) before declaring done; disable with CLAUDE_TQ_INTENT_GATE=0. Open-questions tracker: when the model leaves an answer-worthy question hanging it records it as a native '❓ ...' task; on the NEXT prompt the capture hook re-surfaces any unanswered ❓ (even on a trivial or paused prompt) so a new prompt doesn't bury it, and hud shows an ambient ❓N count — the model re-raises and marks each done once answered or dropped (disable with CLAUDE_TQ_OPEN_Q=0). A per-repo pause (tq-pause.sh) suppresses both the review loop and intent capture on request, so substantive prompts run straight in auto; opt-in agent-mode (tq-agent.sh) fans independent, low-blast-radius tasks out to subagents. Read-only over ~/.claude/tasks (it writes only its own cache state: pause/agent flags + the intent of record).
Generate filled Bug Report and WI (Work Item) completion writeups for Notion, auto-filled from the work done with commits and PRs
Task Trellis Teams - Agent Teams-based hierarchical task management for AI coding agents
from alejandrob-dev/claude-tasks-management
Tatari-specific QA automation for Claude Code - Playwright-powered browser testing for Tatari2 workflows
Tauri 2 setup, migration, config security, IPC, plugins, debugging, testing, and distribution workflows.
Channels Taylor Otwell, fighting Laravel over-engineering.
Test-driven development workflow (red-green-refactor, Storybook/Vitest/E2E guidance)
Test-Driven Development plugin with strict Red-Green-Refactor enforcement, testability auditing, and multi-language test implementation.
Test-Driven Development methodology: Kent Beck TDD, Flutter TDD, and CI triage skills for Claude Code agents.