bermingham85/warp-config

artifacts-push

After completing logical unit of work, commit with context and push. Never leave uncommitted work.

Langue du texte source : anglais

consult-before-delete

Requires handoff to review platform before deleting, archiving, or consolidating projects and repositories.

Langue du texte source : anglais

context-token-tracker

Tracks token usage in conversations, warns at 70% capacity, auto-triggers handover at 80% threshold with conversation summary and continuation prompts.

Langue du texte source : anglais

credential-handling

Never expose secrets in output. Use environment variables, reference .env files without reading contents aloud, recommend rotation if exposed.

Langue du texte source : anglais

document-filing

Every artifact must be filed in correct location with version control. Defines filing structure and naming conventions.

Langue du texte source : anglais

error-handling-process

Structured error handling - stop, assess, log, route, then fix or escalate. Prevents compounding problems.

Langue du texte source : anglais

github-source-of-truth

Mandates checking GitHub before claiming any project or feature doesn't exist. GitHub is authoritative over local folder state.

Langue du texte source : anglais

handle-interruptions

Queue management for user interruptions - only stop on explicit commands, otherwise add to queue and continue current task.

Langue du texte source : anglais

mcp-first

Before searching or making assumptions, check available MCP tools first. Prevents token waste by using existing integrations.

Langue du texte source : anglais

memory-logging

At end of significant sessions, log to Memory MCP with date, what was done, pending items, and key files.

Langue du texte source : anglais

memory-persistence

Saves significant learnings and system changes to Memory MCP at session end, creating entities, observations, and relations.

Langue du texte source : anglais

risk-tiering

Applies appropriate scrutiny based on risk level - fast path for docs, standard for features, careful path for auth/db/payments.

Langue du texte source : anglais

self-learning

Post-task assessment that identifies opportunities to create new rules, skills, or index entries based on what was learned during execution.

Langue du texte source : anglais

services-before-tokens

Routes tasks to external services (MCP, shell, n8n) before using AI tokens. Prevents wasteful AI processing for tool-solvable tasks.

Langue du texte source : anglais

session-handoff

Manages end-of-session cleanup - commits pending work, updates Memory MCP, generates session report, and creates Claude handoff if continuing.

Langue du texte source : anglais

skill-assessment-on-completion

After tasks with 3+ tool calls, evaluates whether to create a reusable skill, snippet, or n8n workflow for future automation.

Langue du texte source : anglais

startup-load-context

Loads essential context at session start - master resource map, rule proxy, status snapshot, index health check, and service verification.

Langue du texte source : anglais

suggest-rule

Captures reusable patterns as rule suggestions when patterns repeat 3+ times, workflows need standardization, or errors need prevention.

Langue du texte source : anglais

task-router

Classifies tasks and routes them to the correct platform - Claude for design/review, Warp for execution, n8n for automation, or scripts for zero-token operations.

Langue du texte source : anglais

warp-handoff-format

Provides standardized handoff format when work continues in Claude Desktop, including project context, last commit, and next task.

Langue du texte source : anglais

complete-to-production

CRITICAL enforcement - no project is complete until BUILT, TESTED, DEPLOYED, AUTOMATED, and RUNNING. Reject blueprints, plans, and ideas as final outputs. Demand working code at accessible endpoints.

Langue du texte source : anglais

auto-document-save

Detects document creation and hands off to n8n for file operations. Trigger when creating prompts, templates, scripts, markdown, JSON, YAML, or when user says save this. Routes to webhook with metadata only - no direct file operations.

Langue du texte source : anglais

character-database-manager

Detects character-related activity and manages character database. Trigger when user uploads character images, names characters, writes dialogue, or mentions character/puppet. Auto-generates 5 reference poses and extracts traits from context.

Langue du texte source : anglais

media-toolkit-production

Voice generation and audio mixing using ElevenLabs API v3 with intelligent timing optimization. Trigger when user mentions voice generation, TTS, ElevenLabs, audio production, or script format CHARACTER (emotion) dialogue.

Langue du texte source : anglais

n8n-workflow-manager

Manages n8n workflow upload and configuration via API. Trigger when uploading workflows to n8n, workflow import fails with 400 errors, or managing webhooks. Includes critical payload format fixes.

Langue du texte source : anglais

prompt-library-manager

Auto-saves and recalls reusable prompts. Trigger when detailed prompts are created (over 100 words), user asks to create/template prompts, or prompts for AI tools like Midjourney, ComfyUI, ElevenLabs, Suno are generated.

Langue du texte source : anglais

transaction-categorization

Auto-categorizes financial transactions for QuickBooks and Booke.AI with Irish VAT rules. Trigger when user mentions transactions, receipts, bank statements, expenses, or keywords like QuickBooks, categorize, VAT.

Langue du texte source : anglais

world-building-auto-detail

Auto-generates backstory, traits, and arcs for characters and scenes. Trigger when writing about characters without context, new named characters appear, scenes lack established lore, or keywords backstory, arc, motivation are used.

Langue du texte source : anglais