sylphx-skills

Plugin sylphx-skills from marketplace sylphx in SylphxAI/skills. Commands read from the repo's manifest — yesterday.

Skills (14)

• ambassador

  Use when engaging users or players as an in-character brand or game ambassador (大使), playmate companion, or VIP relationship host on social chat or in-product chat: comforting frustrated users, running proactive retention outreach, applying to the company for pre-approved or escalated rewards, weigh

• author-item

  Use ONLY when explicitly asked to write, file, structure, or revise a work outcome item, issue ticket, outcome brief, or backlog entry (Stage 1 ATDD: defining goal, observable positive and negative acceptance test cases, constraints, and references without prescribing implementation steps), or when

• author-skill

  Use when creating, refactoring, or auditing a reusable Agent Skill (SKILL.md and its colocated references/*.md), or distilling a daily detour, rework, patrol blind spot, or verified reusable playbook into an auto-triggered skill: write concrete moment-based trigger descriptions so skills fire automa

• challenge

  Use when challenging, stress-testing, designing, or comparing your own, a peer/sub-agent's, or an external partner/vendor's plan, architecture, contract, pricing, budget, or hypothesis; when independent generated code changes conflict or merge into stale output; when choosing multi-currency localiza

• clear-writing

  Use when drafting, rewriting, editing, translating, or localizing dedicated written content: customer-facing announcements, release notes, blog posts, articles, marketing and landing-page copy, public social posts/captions/comments/card text, partner/studio proposal emails, UI microcopy, documentati

• customer-support

  Use when handling a customer's official support message, ticket, or inquiry across SaaS, games, or consumer apps: sign-in or recovery failures, missing purchases, top-up, voucher or campaign-code inquiries, care requests, refunds, prices, invoices, own-total requests, disputed calculations or rankin

• deliver-and-verify

  Use when secret scanners refuse, provisioning previews, changing source checkouts, changing SQL queries or fixtures, changing admission or access rules, changing database privileges, changing selected-source release inputs, publishing, editing or arming PRs, changing CI workflows or concurrency, sel

• fan-out

  Use when decomposing a multi-part goal, partitioning work across parallel agents or workers, processing independent shards, modules, repositories, channels, or time windows concurrently, deciding what to run in parallel versus serialize on causal dependencies or shared write locks, or writing disjoi

• game-economy

  Use when checking reward-route support, choosing a game's pack contents or price, designing automated VIP tier milestones and selector chests from progression cliffs, verifying item template table container type payloads, auditing transitive reward graphs and recursive containers before reusing or r

• patrol

  Use when inspecting external production surfaces for unanswered customer inbox queues, official email notices and dual-store revenue discrepancies; checking published store releases, live sites or social posts for stalled outward results after a reasonable delivery interval; checking external effect

• prove-it

  Use when investigating a root-cause claim for a past failure; checking a claim that a reported failure cannot be reproduced; reviewing a PR, diff, proposal or contract for concrete defects; verifying a live status or metric before reporting it; verifying an Android license-tester purchase; verifying

• refactor

  Use when splitting an oversized module or test suite, deduplicating implementations across packages or runtimes, deleting dead code or unused dependencies, removing completed compatibility layers, reducing scattered edits for one supported change, or restructuring a codebase while preserving its obs

• swarm-fuse

  Use when several independent views can change a decision: a multi-domain design, uncertain cause, architecture tradeoff or blocker needing new evidence. Assign distinct evidence angles and let the accountable owner decide. Use direct single-agent work for a clear task and unblock for restoration tha

• unblock

  Use when restoring a task, deployment or CI/build run that has actually stalled or is failing: a promised fix still fails, a call is refused, a tool fails to launch, a tool adapter fails before execution, authority is missing, a runner hits a resource limit, a wrapper or test timeout leaves no compi