Deploy the Interactor backend service with AWS SSO and SAM. Use when Zack asks to "deploy backend", "deploy interactor service", "deploy service", run deploy_all_regions.bat, or deploy interactor-service to dev, uat, preprod, or prod; checks AWS SSO for the deployer profile, runs login if needed, then runs the backend deployment script and returns full raw command output without summarizing.

Merge and deploy a reviewed Interactor feature to production with sanity-check gates, Cookieball release logging, changelog capture, and rollback instructions. Use when Zack says "deploy feature P3", "deploy-feature P10", asks Panda, Hammer, or Cookieball to deploy a reviewed feature tag, or otherwise explicitly asks to merge a feature PR to main, deploy backend dev then prod, promote frontend main to dev then prod, sanity check Interactor public pages, and rollback prod backend or UI if prod sanity checks fail.

Generate Interactor training plan Python scripts through an organization-level workflow spanning Dev Team, Marketing Team, Fitness Specialist/Button, Training Plan Image Sketcher/Kitty, and Panda coordination. Use this skill whenever the user asks to generate workout plans, create training plans, make new training routines, research plan ideas, or produce training plan code/images/logs for the platform. Also trigger when the user says "generate some plans", "new plans", "more workout plans", or references the exercise library in the context of creating plans. Generated plans should prioritize fitness quality while using tasteful SEO-friendly names and descriptions, current marketing/research context, community signals, engagement metrics, and generated images.

Review, audit, and improve Interactor's existing exercise library metadata. Use this skill whenever the user asks to go through existing/default/system exercises, clean up exercise names, descriptions, units, tags, or YouTube links, iterate on exercise quality, normalize the hardcoded exercise catalog, or adapt the exercise library to a changed tagging taxonomy.

Use this skill whenever Zack wants to view, summarize, or update his personal loans / debt tracker. Trigger phrases include "summarize my loans", "update my loans", "loan status", "what loans do I have", "show my loans", "loans summary", "my debt summary", "how much am I paying on loans", "when do my loans finish", "which loan should I pay off". Also trigger when Zack provides new loan data (e.g. "just paid X/60 instalment", "outstanding is now S$X", "new monthly is X") for any tracked loan, or shares a screenshot of a loan/instalment statement. The canonical tracker is at `C:\Developer\PersonalNotes\Finance\Loans.md`. Note: in Zack's vocabulary "loans" includes ALL instalment plans — bank personal loans, credit-card cash-on-instalment, device instalments (Apple etc.), renovation loans, and IRAS income-tax GIRO.

Run deep self-learning research for one or more zstack organization agents, persist the full learning in each agent workspace, and assimilate durable takeaways into each agent's role definition and relevant context. Use when Zack asks agents to self-improve, research their specialization, update their expertise, deepen organization knowledge, learn industry standards or best practices, update agent role definitions from learning, or persist research learnings into each agent's own workspace.

Run a recurring zstack organization-wide sync across agents, teams, active Interactor workstreams, product-signal tracking, Google Ads/marketing context, decisions, and follow-up owners. Use when Zack asks for "organization sync up", "org sync", "weekly sync", "sync the agents", "sync dev team and marketing team", "before strategic adjustments", "before marketing plan changes", or wants Panda/Hammer/Picnic/Cheesecake or other agents aligned on current status and implications.

Consolidate Zack's Obsidian Panda mobile notes and turn raw thoughts, reminders, ideas, and zstack candidates into structured actions or durable zstack context. Use when Zack says things like "Panda, consolidate my mobile notes", "consolidate my Panda notes into zstack", "clear my mobile notes inbox", or asks to classify/persist notes from C:\Developer\PersonalNotes\panda.

Use when Zack asks Codex/Panda to remember, rmb, save, add to context, persist, or update durable Markdown memory. Default Codex/Panda destination is zstack Markdown under C:\Developer\zstack\personal\, C:\Developer\zstack\organization\, or C:\Developer\zstack\skills\. Only write to Claude/CLAUDE.md when Zack explicitly names Claude, CLAUDE.md, or Claude Code context.

Use this skill ONLY when the user explicitly asks to "refresh context", "refresh", "refresh yourself", "update yourself on the codebase", "re-read everything", or similar manual triggers. Re-reads all durable context layers (project CLAUDE.md, zstack/personal/, all auto-memory files), pulls recent git activity from active Interactor repos, compares against existing auto-memory, and updates stale entries to match current reality. Slow and thorough by design — meant for the start of a long session after a gap or after big changes have landed. Do NOT trigger automatically at session start, before tasks, or from inside other skills; wait for explicit user invocation.

Research trending, popular, seasonal, niche, or social-media-relevant fitness topics, then generate Interactor training plans through the organization-level generate-workout-plans workflow. Use this skill whenever the user asks to "research trending training plans and generate", "create top trending plans", "find trending plans to create", "generate plans based on what's popular", "make plans for the current season", "what's trending in fitness, make plans", or any variation about researching current trends / popularity / seasonal relevance and producing plan files, images, and research/crafting logs.

Prepare an existing PR feature branch for Zack's code review. Use when Zack says "review feature", "review PR", "I want to start reviewing P#", "start reviewing P10", or otherwise asks to review a feature that should resolve to an existing GitHub PR; checks out the PR branch in the main local repo, merges latest main, resolves conflicts, validates, pushes the feature branch, and does the same in both Interactor backend and frontend when the feature spans both repos.

Use when creating, updating, reviewing, or validating zstack Codex skills, SKILL.md files, skill registry entries, trigger descriptions, skill junctions, or reusable workflow templates.

Lightweight zstack context reload for cross-chat synchronization. Use when Zack says "sync chat", "sync chats", "light refresh", "reload zstack context", "refresh zstack context", "sync from zstack", "reload memory", "reload org context", "read latest zstack", or asks an older chat to pick up zstack changes made by another chat; runs a fast curated zstack pull/read workflow rather than the heavier refresh-context skill.

Use when investigating bugs, failing tests, build errors, regressions, flaky behavior, performance issues, or unexpected Interactor behavior before proposing fixes; especially when root cause, evidence, and verification matter.

Use when Zack asks Panda, Picnic, Cheesecake, Dev Team, or the organization to run, create, improve, or refer to the weekly Interactor business review covering Google Ads, marketing performance, engagement metrics, community signals, feature priorities, business direction, pivots, and process improvements.

Run a read-only health dashboard or doctor/validator for Zack's zstack operating memory and AI-agent organization. Use when Zack asks for "zstack health", "zstack doctor", "health check zstack", "score zstack", "zstack scorecard", "context hygiene", "is zstack healthy", "check zstack organization", "validate zstack", or wants to find stale TODOs, missing agent sections, broken zstack paths, skill registry/junction issues, or sensitive-info risks.

Understanding Level: U2 Use this file when one agent delegates work to another, when multiple agents collaborate, or when a task needs to be resumable later. Keep handoffs compact...

Use this template as a starting point, then remove sections that do not help the agent act correctly. markdown name: example skill name description: Use when [specific request sha...