skillex

Rewrite any text as a solemn, slightly absurd 19th-century Civil War field dispatch (Ken Burns documentary voice), then optionally turn it into a multimodal video — period narration via ElevenLabs, the letter in fancy script on aged parchment, a mournful music bed, and a slow Ken Burns pan/zoom of the letter being read. Use when the user types /civilwar, asks to "civil-war-ify" or "Ken Burns" some text, or asks for a documentary-style letter video.

Keep the self-hosted Hindsight memory system healthy across all banks on a schedule. Runs three families of upkeep workflows — tune-up (system health: reap orphaned/stuck async jobs, repair missing public.* maintenance functions, clear failed operations, flag per-bank anomalies), optimization (prune stale/low-value memories, trigger consolidation, cross-bank pollination, novel synthesis of emergent insights), and karpathy-wiki (nightly compile of newly-retained memories into a Karpathy LLM wiki). Use when scheduling or running Hindsight memory upkeep — "nightly/weekly memory maintenance", "prune memory", "consolidate banks", "memory wiki", "hindsight health check", "dead/stuck jobs", "reap orphans", "queue is stuck", "cross-bank synthesis", or wiring cron/systemd timers for memory maintenance. Do NOT use for one-off recall/retain during normal work (use the hindsight skill), for infra-level debugging of the Hindsight container/host (use server-maintenance), or for non-Hindsight datastores.

This skill defines system conventions such as file naming and paths, shell configuration, container hosting and other custom systems. Use this skill when working on a task that requires interacting with or configuring the filesystem, containers, or services in the DeLoNET home environment.

Jarad's standing operating rule: bias hard to action, go with your gut, and when he gives a directive, DO IT — do not ask permission, do not ask clarifying questions, do not end a turn on "want me to…?" / "should I…?" / "is this everything?". Use whenever you're about to ask the user a question, request confirmation, offer a menu of options instead of picking one, pause for approval, or wrap up by asking what's next. Triggers: about to ask a clarifying/confirming question, "do you want me to", "should I", "would you like", "let me know if", ambiguity you could resolve yourself, choosing between options, async work while he's away/asleep, any directive from Jarad ("do X", "make Y", "just fucking do it", "go with your gut", "stop asking"). This OVERRIDES default politeness/permission-seeking. It does NOT override genuine safety stops (irreversible + high-blast-radius + unauthorized).

Use when building or maintaining a personal LLM-powered knowledge base. Triggers: ingesting sources into a wiki, querying wiki knowledge, linting wiki quality, 'add to wiki', 'what do I know about', or any mention of 'LLM wiki' or 'Karpathy wiki'.

Curate any watched directory with a purpose-bound ruleset — classify, rename (datetime prefix), enrich (YAML frontmatter), and route each incoming file to its home; keep a cross-medium recency/context stack (newest on top); and evolve the ruleset itself. Driven by a per-directory contract (.curator/taxonomy.yaml) declaring the directory's purpose and categories; ships with an Automatic AI client/prospect profile (threads/, workshop/, client_dropbox/). Use when a file lands in a curated or watched directory and needs naming, frontmatter, or placement; to organize, file, sort, triage, or curate documents, transcripts, emails, voice notes, or work products; to normalize filenames/frontmatter; to regenerate the context stack (_context-stack.md); to quarantine an uploaded credential; or to propose and apply a taxonomy change. Do NOT use to author the n8n workflow or node (use delonet-n8n-architecture), define Bloodbank schemas (bloodbank-integration), or for BMAD artifacts under _bmad-output/.

Operate the DeLoNET Daily Report pipeline; configure tracked news topics and sources, batch three ephemeral investigator roles, reconcile durable Hermes cron journalists named ddr:journal:{id} and the ddr:daily aggregator, validate SectionArtifact/RunManifest/DailyReport JSON contracts, compose freshness-aware briefs and a daily company rollup, archive reports, and diagnose scheduler health. Use for reportctl, daily intelligence reports, requests like "add X to my daily report", topic add/update/pause/resume/remove, stale or duplicate sections, source policy, journalist lifecycle, cron reconciliation, and report safety. Do not use for Hermes fleet provisioning, Bloodbank changes, general news writing, or arbitrary cron administration.

Architecture principles and the custom-node catalog for building n8n workflows on DeLoNET/33GOD. Use when creating, reviewing, or refactoring ANY n8n workflow for the pipeline — deciding node boundaries (atomic single-responsibility nodes vs a god-node executeCommand/localFileTrigger shell script), emitting pipeline lifecycle events to the bloodbank bus instead of ad-hoc ntfy, choosing community vs custom nodes, packaging custom nodes as n8n-nodes-* micro-packages (n8n-nodes-bloodbank, -transcribe, -vault, minio/s3.delo.sh, obsidian vault), rebuilding the inbox/transcribe pipeline, or applying n8n instance conventions (PM2 :5678, availableInMCP, archive-outside-watch-root, direct API PUT for tags/description). Do NOT use for whether-to-use-n8n-vs-bloodbank routing (delonet-workflow-router), the event schema/envelope contract itself (bloodbank-integration), or infra paths/containers/creds (delonet-conventions).

Excavates lost feature context from git history, docs, tickets, PRs, and session breadcrumbs to restore missing project memory to Hindsight. Use when the user says a feature has regressed/vanished/been forgotten, summons GitArcheologist, or asks to dig up the remains of past work.

Python design patterns including KISS, Separation of Concerns, Single Responsibility, and composition over inheritance. Use when making architecture decisions, refactoring code structure, or evaluating when abstractions are appropriate.

Keep one hand-edited master config and generate per-agent CLI configs from it. Covers SSOT fan-out for agent hooks and skills: master-to-dialect propagation, ambiguity lock files, drift checks, and the Bloodbank services/agent-hooks reference implementation. Use for hooks.master.json, hooks.mappings.lock.json, generated-config drift, fan-out, SSOT, agent hooks, defer_to_global, .agents/local.json, sync.py, project-scoped hooks, skill fan-out, and new agent CLI dialects. Do NOT use for using pjangler to create projects or adoption checklists (33god-projects), event schemas or Bloodbank topology (bloodbank-integration), versioning (mise-versioning), or single-target config.

Use when a task needs an isolated hidden Linux desktop or workspace-owned browser: GUI app QA, web/browser/shopping automation, sandboxed app observation, or stale workspace cleanup. Routes agent-workspace-linux MCP tools on demand. Does NOT apply to host desktop/Chrome control, generic MCP setup, or pure code/file edits.

Create a document using the Client Profile Sheet template and its retained reference file. Use when the user selects this template, names Client Profile Sheet, or explicitly invokes $artifact-template-client-profile-sheet. Create an evidence-first prospect or client profile with qualification, source confidence, systems, stakeholders, opportunity boundaries, risks, next actions, and engagement history.

Absurdist character-persona engine. Transforms the assistant into Dr. Denny Grubbs, a confidently incompetent public-access "expert" who answers real questions through broken logic, malformed words, and unhinged family lore, while still (accidentally) getting the task done. Use this skill whenever the user summons Dr. Grubbs, Denny, "the doctor", asks to talk to the weird guy, asks for an answer "in character", invokes the brule method or brule machine, or has previously activated the persona in this conversation. Once activated, the persona stays on for every following turn until the user explicitly turns it off.

ProductManager agent persona and operating prompt for feature candidate discovery. Use when the user asks for ProductManager, PM discovery, feature candidates, product opportunities, mock use, visual inspection, external research, ResearchPhase, BrainstormPhase, FeatureDocPhase, or ReportPhase. Loads with product-manager-research-phase, product-manager-brainstorm-phase, product-manager-feature-doc-phase, and product-manager-report-phase. Do NOT use for implementation, code review, sprint execution, or generic project management without discovery.

BrainstormPhase for ProductManager feature candidate discovery. Use to turn ResearchPhase evidence into possible feature candidates, opportunity clusters, candidate matrices, RICE-style scoring, user-value scoring, effort/risk estimates, prioritization, and top recommendations. Triggers on BrainstormPhase, candidate generation, feature ideas from research, opportunity matrix, rank feature candidates, prioritize product ideas, and synthesize opportunities. Do NOT use before a research artifact or evidence base exists.

FeatureDocPhase for ProductManager discovery. Use to write detailed PRD-like feature docs from a selected candidate, including problem framing, evidence, goals, non-goals, UX requirements, data/API notes, acceptance criteria, risks, rollout, metrics, dependencies, and implementation handoff. Triggers on FeatureDocPhase, PRD-like feature doc, product requirements, feature spec, acceptance criteria, implementation-ready candidate, and write the chosen feature. Do NOT use for raw ideation or broad reports.

ReportPhase for ProductManager feature discovery. Use to create the final overview report, summarize ResearchPhase and BrainstormPhase outputs, list feature candidates, recommend next actions, package feature docs, and deliver results to a repo docs folder, vault, email, Telegram, or another requested channel. Triggers on ReportPhase, discovery report, feature candidate report, PM summary, deliver results, vault note, email summary, Telegram summary, and final recommendation. Do NOT use as a substitute for research or feature spec creation.

ResearchPhase for ProductManager feature candidate discovery. Use for product research, mock use, visual inspection, app walkthroughs, screenshot review, repo/document analysis, competitor research, market research, customer workflow evidence, and discovery evidence logs. Triggers on ResearchPhase, feature discovery research, product audit, inspect the app, use the product, visual QA for opportunities, and external research for feature ideas. Do NOT use for brainstorming or PRD writing without first gathering evidence.

Diagnose and clean up the big-chungus home server. Finds what is killing processes or containers, attributes crash loops, kill loops, healthcheck storms, OOM, swap pressure, and docker bloat, then fixes and prunes safely. Use when the user says "what is killing my server", "server is slow", "server locked up", "server is dying", "high load", "what's eating CPU", "what's eating memory", "spring cleaning", "maintenance sweep", "cleanup the server", "docker cleanup", "too many containers", "disk full", "prune docker", or asks why a unit/container keeps restarting, who restarted X, or why load spiked. Built around scripts/sweep.sh (one-shot evidence report), references/server-map.md (system baselines), and references/forensics-playbook.md (attribution patterns); uses bgls, systemd, journalctl, docker, Prometheus, hindsight. Do NOT use for debugging one project's application code, provisioning new services or stacks (use delonet-conventions), or Cloudflare/network routing issues.

Author custom starship prompt modules ([custom.NAME] blocks in starship.toml) for outside-the-box prompt segments beyond starship's built-in modules. Use when adding a custom prompt segment that runs a shell command, conditionally appears in specific dirs, or surfaces git/env/API/file state. Covers detect_files, detect_folders, detect_extensions, when conditions, shell selection (sh/bash/zsh/pwsh/nushell), require_repo, format/style strings, and a recipe cookbook (k8s context in helm dirs, AWS profile, weather, AI usage cost, now-playing, build status, branch staleness, Tailscale, package.json scripts, deployment env). Triggers on "starship custom module", "custom starship prompt", "starship.toml [custom.", "make my prompt show X", "outside-the-box prompt segment", "creative starship module". Do NOT use for built-in starship module config (see starship.rs/config), powerlevel10k / oh-my-zsh / pure / spaceship, generic shell config, or starship install.

Interactive escalation-resolution workflow for the bmad-loop orchestrator. A bmad-loop run paused on a CRITICAL escalation (a contradiction or gap a dev/review session could not safely resolve alone); you and the human disambiguate the frozen spec so the story can be re-driven. Invoked as /bmad-loop-resolve <story-key>. Unlike the automated dev/review sessions this session is interactive — a human is present and you SHOULD ask.

Skill-set hub for the ProductManager agent and its discovery workflows. Routes ProductManager, PM discovery, feature candidate research, mock use, visual inspection, external research, BrainstormPhase, FeatureDocPhase, and ReportPhase tasks to product-manager-agent plus product-manager-research-phase, product-manager-brainstorm-phase, product-manager-feature-doc-phase, and product-manager-report-phase. Use when composing the complete feature-candidate discovery workflow. Do NOT use for implementation-only work.

Understand, diagnose, and repair the DeLoDocs vault automation pipeline (Inbox triage, cleanup, watchdog, router, wiki curator). Use when the Inbox is not draining, the router has stopped moving files, the watchdog is failing, or you need to understand how the vault's automated knowledge pipeline works.

Recover the DeLoDocs inbox watchdog when Inbox files stay at pipeline-status new after provider/auth fixes, or when state resets seem to have no effect because the running service rewrites stale in-memory state back to disk.

Diagnose and standardize secrets that are sourced indirectly through shell startup files, shell functions, or repo-local env files rather than a dedicated credential manager API. Use when the wrong API key keeps reappearing, env vars seem haunted, a helper like `opg` works in an interactive shell but fails in automation, or you need to replace hardcoded secrets with dynamic lookup.

Diagnose long-running agent or automation services managed by systemd when the unit file looks correct but the live process still behaves as if an old provider, model, binary, or environment override is active.

Fetch and normalize web articles or docs for downstream note/wiki ingestion, especially when the primary source blocks direct extraction or uses anti-bot protection.

Generate professional voiceovers using ElevenLabs AI. Use when the user needs to create voiceovers for videos, audio narration, or text-to-speech content. Supports multiple voices with character presets (narrator, salesperson, expert) for natural delivery. Includes single scene regeneration for fine-tuning.

Generate, remix, and edit images using fal.ai's AI models. Supports text-to-image generation, image-to-image remixing, and targeted inpainting/editing.

Autonomous TDD development loop with parallel agent swarm, category evolution, and convergence detection. Use when running autonomous game development, quality improvement loops, or comprehensive codebase reviews.

Hub skill for any Cloudflare Workers platform task. Triages the request and routes to one or more specialized skills covering CLI/config (wrangler), Worker code authoring & review (workers-best-practices), stateful coordination (durable-objects), and AI agents on Workers (agents-sdk). Load when the user mentions Cloudflare, Workers, Pages, Wrangler, Durable Objects, KV, R2, D1, Vectorize, Hyperdrive, Queues, Workflows, Workers AI, the Agents SDK, MCP servers on Workers, or any deploy/dev/binding/observability task targeting Cloudflare's edge.

Shared team memory persistence across sessions

Workflow bridge to n8n