3D-Stories

WF14 — assess the WORKFLOW MACHINERY of a just-completed rawgentic run (WF1/WF2/WF3/WF5/WF13/epic driver) and route feedback to rawgentic development. Use after any completed WFn run — the user says "assess the workflow run", "run feedback", "post-run assessment", "how did the workflow itself do", or an embedding workflow invokes it with explicit args. Do NOT use to review the deliverable the run shipped (that is WF5 / code review), to assess non-rawgentic workflows, or to fix any defect it finds (report-only).

WF17 — mine session history for recurring skill/command candidates (detect → queue → synthesize → gate). Use when the user asks to mine sessions for patterns, "what keeps recurring", "what skills should we build", after a campaign wraps, or on-demand workflow-improvement hunts. Report-only — writes only the candidates queue and a report pair; accepted candidates route to WF1 as prepared drafts, nothing is ever auto-filed. Invoke with /rawgentic:session-mining.

Full-text search over past Claude Code session history (all projects) via the local FTS5 session index. Use when you need to find what a past session did, said, or decided — "what did we do about X", "which session touched Y", "when did we discuss Z", "search my session history" — and mempalace recall returns nothing or you need the raw conversation text rather than curated memories. Read-only over a local, derived, rebuildable index; never egresses session content. Invoke with /rawgentic:session-recall followed by a search query.

Implement a feature (or a design-heavy/complex bug fix) from a GitHub issue through the WF2 16-step workflow with TDD, multi-agent code review, quality gates, and — when the project configures them — CI and deployment verification. Invoke with /implement-feature followed by a GitHub issue number or URL. For a narrow, reproducible bug fix prefer /rawgentic:fix-bug (WF3); implement-feature is the home for features and for bugs that need full design + implementation. Only trigger when the user explicitly invokes /implement-feature or /rawgentic:implement-feature.

Open and file a NEW GitHub issue — a feature request or bug report — for the active rawgentic project. Use whenever the user wants to capture a desired feature/enhancement or an observed/reproducible bug as a tracked issue, however phrased ("open/log/raise/file an issue", "write up a bug report", "file a feature request", "put it on github", "track this", "users keep asking for X"), even when no repo is named. It targets the repo from the project config, checks for duplicates, conforms to the issue template, and verifies referenced code exists. Do NOT use to implement/fix/code the change itself, to list/search/read existing issues, to comment on or review a PR against an issue, or to edit issue-template files. Invoke with /create-issue followed by a description of the desired feature or observed bug.

Fix a bug using the WF3 14-step workflow with reproduce-first TDD, root cause analysis, lightweight reflect, and conventional commit PR. Invoke with /fix-bug followed by an issue number. Only trigger when the user explicitly invokes /fix-bug or /rawgentic:fix-bug.

Configure a project's .rawgentic.json — the structured config that all rawgentic workflow skills depend on. Auto-detects tech stack, testing, CI, database, services, and more for existing codebases; brainstorms intent for blank projects. Handles migration from old CLAUDE.md-based rawgentic setups. Use this whenever a project needs initial configuration, reconfiguration, or when the session-start hook says "Config missing -- run /rawgentic:setup."

Register a new or existing project in the rawgentic workspace. Creates the workspace file on first run, handles GitHub cloning or git init for new folders, and delegates to /rawgentic:setup for configuration. Use when starting a new project, adding an existing project to your workspace, or when the session-start hook says "Run /rawgentic:new-project to get started."

Persistent virtual display + VNC for headless servers. Four commands: /vnc-service:setup (install and configure), /vnc-service:run (ensure running + print connection info), /vnc-service:stop (stop services to free resources), /vnc-service:status (check health). Requires Debian/Ubuntu, systemd, sudo. VNC access restricted to local network via ufw firewall. Use whenever browser interaction is needed on a headless server — OAuth, CAPTCHA, 2FA. Other skills should call /vnc-service:setup as a prerequisite check and /vnc-service:run before browser operations.

Set up a persistent virtual display + VNC server on a headless Linux server. Installs Xvfb, x11vnc, creates systemd user services, sets a VNC password, and configures display :99. Requires Debian/Ubuntu (uses apt-get). Use this skill when a tool needs browser interaction on a headless server — OAuth login, CAPTCHA solving, 2FA prompts. Trigger on: "vnc setup", "set up vnc", "headless browser", "no display", "DISPLAY not set", "CAPTCHA requires browser", "OAuth needs browser", or when any other skill calls /vnc-service:setup as a prerequisite.

Ensure the persistent virtual display + VNC is running and print connection info. Call this skill whenever a browser interaction is needed on a headless server — before OAuth login, CAPTCHA solve, or 2FA prompt. If the service isn't set up yet, directs to /vnc-service:setup. Trigger on: "start vnc", "vnc run", "I need to solve a CAPTCHA", "browser login needed", "OAuth prompt", or when any other skill needs browser interaction on a headless server.

Show the status of the virtual display and VNC service — whether running, display number, port, and connection info. Use when checking if VNC is available or troubleshooting display issues. Trigger on: "vnc status", "is vnc running", "display status", "check vnc".

Stop the virtual display and VNC server services. Use when you no longer need browser interaction and want to free resources. The services can be restarted anytime with /vnc-service:run. Trigger on: "stop vnc", "vnc stop", "disable vnc", "turn off vnc", "shut down display".

Search long-term memory, invalidate stale decisions, view decision timelines, or browse cross-project tunnels. Supports subcommands: search (default), invalidate, timeline, tunnels.

Upgrade the mempalace dependency to the latest version. Displays old and new version numbers, warns on major version changes.

Start, stop, or check status of the web frontend containers for browsing memory data.

Set up the Replicate MCP server for image generation. Detects and installs prerequisites (Node.js, npm, npx), configures the API token, and registers the MCP server with Claude Code. Use when user says "setup replicate", "configure image generation", when generate-image skill reports missing MCP, or when the user gets errors trying to generate images. Also use proactively if you detect Replicate MCP tools are missing when the user asks to generate an image.

Generate images for websites, presentations, and marketing materials using AI models on Replicate (Nano Banana 2, FLUX Schnell) with optional background removal for transparent PNGs. Use this skill whenever the user asks to create, generate, or make images, illustrations, icons, hero images, OG images, logos, favicons, or any visual asset. Also use when implementing issues that require visual assets, or when the user says "generate an image", "create an illustration", "make me an icon", "I need a visual for...", "remove background", "make transparent", or references image generation in any context. If a project needs visual assets and you're about to use placeholder text/emoji instead, use this skill to generate real assets.

Comprehensive competitor research pipeline that consolidates scattered sources into curated market briefs. Pulls existing NotebookLM sources, searches the web via Google AI Mode MCP, verifies accuracy, and produces structured per-section markdown files plus a combined brief for NotebookLM upload. Use this skill whenever the user wants to research a competitor, build a competitor profile, create a market brief, consolidate competitor sources, audit competitive intelligence, or mentions "competitor research", "market brief", "competitive analysis", "competitor profile", "competitive landscape", or "market intelligence". Also trigger when the user asks to compare their product to another company, research a specific company (e.g., "research Greenlight", "find out about Finch"), asks "who are our competitors", wants to analyze the competition, or asks to clean up or consolidate NotebookLM sources about a specific company. Includes a setup mode (/competitor-research setup) for first-run dependency configuration.

Comprehensive competitor research pipeline that consolidates scattered sources into curated market briefs. Pulls existing NotebookLM sources, searches the web via Google AI Mode MCP, verifies accuracy, and produces structured per-section markdown files plus a combined brief for NotebookLM upload. Use this skill whenever the user wants to research a competitor, build a competitor profile, create a market brief, consolidate competitor sources, audit competitive intelligence, or mentions "competitor research", "market brief", "competitive analysis", "competitor profile", "competitive landscape", or "market intelligence". Also trigger when the user asks to compare their product to another company, research a specific company (e.g., "research Greenlight", "find out about Finch"), asks "who are our competitors", wants to analyze the competition, or asks to clean up or consolidate NotebookLM sources about a specific company. Includes a setup mode (/competitor-research setup) for first-run dependency configuration.

Use when asked to repair, de-slop, tighten, or edit prose (essays, specs, PRDs, ADRs, READMEs, docs) for editorial quality WITHOUT losing meaning, requirements, uncertainty, or the author's voice. Diagnoses genericness, unsupported claims, synthetic cadence, obscured responsibility, and voice discontinuity, then proposes minimal passage-local repairs. NOT an AI-authorship detector — never strips a stylistic feature just because it looks AI-written. Triggers include "de-slop this", "make this less generic", "tighten this doc", "is this AI-slop", "edit this without flattening my voice", and the audit/suggest/apply commands.

End-to-end presentation creation through structured phases: research, requirements, design, visual generation, implementation, and iteration. Handles PPTX, HTML, PDF, and DOCX output formats via document-skills. Use this skill whenever the user wants to create a presentation, slide deck, pitch deck, talk, or lecture — even if they just say "build me a deck about X" or "I need slides for a talk on Y." Also use when the user has existing content (notes, docs, research) they want turned into a presentation. Triggers on: "create presentation", "build slides", "make a deck", "presentation about", "slide deck for", "put together a talk", "pitch deck", "build me a pptx", or any request that involves creating structured visual content for an audience.

Bidirectional sync between BMAD .md artifacts and Jira Cloud / Confluence Cloud. Syncs stories, epics, and sprint status to Jira issues; publishes sprint pages, retrospectives, and change proposals to Confluence. Use when the user says "sync to jira", "push to atlassian", or "pull from jira".