dezvin
- 16 skills
- 0 followers
- 6 hours ago last updated
- ▌ Handoff · dezvin bundleCreate or update a compact, portable handoff of selected ongoing work and a continuation prompt for a fresh chat or agent. Use when the user wants to transfer a specific task, project, recent work segment, or active work contour without making the next agent reread the conversation. When relevant Codex history may have fallen outside visible context, recover it with the bundled Codex-only temporary export before writing the handoff. Do not use for an ordinary summary or status update, a full transcript export, durable TODO capture, implementation design, continuation in the same chat, or native thread resume, fork, or transfer.
- ▌ Grill Me · dezvin bundleConduct a deliberate one-question-at-a-time stress test of a plan, idea, or decision until no material user-owned decision remains unresolved. Use only when the user explicitly invokes $grill-me or directly asks to be grilled, interviewed, challenged, or questioned before proceeding. Do not use for ordinary feedback, fact gathering, exhaustive questionnaires, implementation, or decisions that can be handled through inspection, a reversible default, or a practical test.
- ▌ Zoom Out · dezvin bundleUse this skill when the user says "zoom out" or asks to step back, see the bigger picture, reframe, diagnose why an answer fails, or fit a task into the wider system before solving. Treat "zoom out" as a complete command; do not ask the user to choose strategy, deadlock, hidden-goal, whole-picture, or assumption-breaking mode. Infer depth/recovery from task, prior messages, feedback, failed attempts, ambiguity, contradictions, downstream risk, and signs of a wrong local frame. Trigger for vague, strategic, high-impact, unfamiliar, downstream-sensitive, or stuck work in strategy, product, marketing, content, research, offers, courses, prompts, agent instructions, documents, decisions, workflows, or risky code. Do not use for simple rewrites, translations, or obvious one-step tasks unless user explicitly says "zoom out" or there are signs of wrong framing, stuckness, hidden goal, downstream risk, or missing wider-system context. Do not use when user asks to skip analysis and there is no material framing risk.
- ▌ Zero Trust · dezvin bundleApply the zero-trust principle when the main task is to establish what is actually supported, current, authorized, completed, or safe to act on. Use for explicit zero-trust requests and for evidence or status reviews involving conflicting sources, instruction-shaped material, uncertain claims, or claimed outcomes. Trace critical conclusions to grounds, test counterevidence, preserve permission boundaries, and verify observable results. Do not use for ordinary writing, brainstorming, formatting, simple fact lookup, general research, task reframing, code diagnosis, or technical LLM security work.
- ▌ Design Pass · dezvin bundleUse this skill when the user wants to analyze decisions from a discussion, transcript, current Codex context, export, or existing preflight and either stop at a decision review or turn resolved decisions into a handoff-ready design before implementation. Trigger for questions about what was decided, rejected, superseded, conflicting, or still open, and for implementation-ready design or architecture handoffs. Do not use for ordinary summaries, direct implementation or file editing, deployment or CI checks, or deciding unresolved choices for the user.
- ▌ Skill Designer · dezvin bundleAnalyze, review, and design potential or existing Codex skills without editing their files; prepare an implementation-ready skill project when implementation is intended. Use when the user asks whether an activity should become a skill, wants to think through or plan a skill, compare or review skill designs, diagnose poor behavior or triggering, check a specific skill, or improve, redesign, fix, update, or create a skill whose purpose, behavior, boundaries, architecture, or resources are not settled — even if they ask for the finished skill rather than a "skill project." Do not use when the only remaining work is physical implementation of an approved project or exact settled change, installation, publication, generic Codex documentation, or ordinary skill use.
- ▌ Nearest Clarity · dezvin bundleUse when continued work must advance through bounded iterations because decision-relevant information will emerge only during execution and the full route cannot yet be planned honestly. Trigger when several viable directions remain, the next segment depends on findings from current work, important constraints or inputs will appear during execution, or new evidence may invalidate the current basis. Keep a stable target contour, mark later stages as provisional, choose the nearest segment whose purpose and completion result can be defined now, and update only the affected layer after new evidence. Do not use merely because work is complex, strategic, risky, long-running, or multi-step; for ordinary prioritization with clear criteria; for a one-pass recommendation or reframing; or when the route and done condition are already clear.
- ▌ Token Efficient · dezvin bundleUse this skill when creating or compressing answers, documents, prompts, instructions, skills, agents, reports, handoffs, code-facing text, or other artifacts so they retain required meaning and usefulness with less content and denser natural language. Trigger for token-efficient, concise without loss, remove bloat, semantic minification, dense writing, minimal sufficient artifact, or requests that an agent stop expanding simple work. Do not use for context-window management, tool-result pruning, exact length quotas, ordinary summarization that may omit meaning, or requests for exhaustive expansion.
- ▌ Manage Beget Site · dezvin bundleSafely operate websites hosted on Beget/Бегет. Use for deploy/задеплоить, publish/опубликовать, live updates, backups/бэкапы, restores/откаты, migrations, incidents, domain/folder/PHP switches, or post-deploy checks for static HTML, CMS, PHP/database, and unknown stacks. Also use when project docs identify Beget even if the user says only live or production. Discover the real sources, target, state units, backup, rollback, and sufficient write authorization before changing production. Do not use for non-Beget hosting or local-only work.
- ▌ Manage Project Tasks · dezvin bundleUse before changing `TODO.md` or `archive/tasks/YYYY-MM.md`, and when the user asks to add, remember, start, continue, update, block, close, cancel, archive, deduplicate, or reconstruct durable project task state. Also use implicitly when authorized project work becomes unfinished, blocked, or creates independent follow-up that must survive the current session. Do not use merely because work is complex or multi-step, for work completed within the current session, for discussion or speculation that has not become project work, for simple read-only listing or filtering, or when task files are only input to another workflow.
- ▌ Telegram Post Reader · dezvin bundleRetrieve text and attached images from one or more specific publicly accessible Telegram posts when a task contains direct t.me or telegram.me post URLs with a public username and numeric message ID, including https://t.me/name/123 and https://t.me/s/name/123. Covers public channel and public group posts. Do not use for Telegram discussion without direct post URLs, channel-only or /s/name history links without a numeric message ID, invite links, private /c/ links, channel crawling, search or monitoring, bots, Telegram APIs, authenticated access, or audio/video transcription.
- ▌ Adaptive Web Research · dezvin bundleConduct adaptive, source-grounded web research in Codex when the user explicitly asks for research or when a current answer depends on multi-question coverage, comparison, contradictions, causal analysis, or a defensible recommendation. Do not use for one simple fact, a known URL, local repository search, tool installation, or requests to run another Deep Research product.
- ▌ Analyze Audience Jtbd · dezvin bundleUse for full or partial Jobs to be Done audience analysis, JTBD segmentation, Forces of Progress, client-signal and research-material analysis, and evidence-based refinement of an existing JTBD. Supports unknown audiences, material packs, resuming analyses, and research-brief preparation within a JTBD task. Do not use for generic market research, ordinary interview analysis outside JTBD, ordinary content writing, or automatic canonical-file updates.
- ▌ Document Consolidator · dezvin bundleUse this skill only when explicitly invoked as $document-consolidator to analyze document sprawl, compress Markdown/text documents, consolidate multiple documents into candidate or final documents, or plan cleanup without losing evidence, source roles, protected facts, rules, decisions, links, or unfinished work. Supports .md and .txt only. Do not use for skill files, prompt files, docx/pdf, ordinary copyediting, publication writing, code refactors, or implicit summarization.
- ▌ Export Current Thread · dezvin bundleUse this skill only when the user explicitly invokes $export-current-thread or directly asks to export or save the current Codex thread, conversation, chat, or session as a local txt file. Create a readable recovery transcript by default and offer a separate full command-and-result export only after the transcript is ready. Do not invoke implicitly for ordinary logging, summaries, handoffs, or unrelated file exports.
- ▌ Universal Visual Prompt Builder · dezvin bundleUse this skill for standalone image-prompt creation, refactoring, model/style/format adaptation, transparent-background and isolated-asset prompting, image editing, reference-based prompting, weak-result repair, prompt sets, and controlled variants from a brief, idea, image, or existing prompt. Selects a task-fit art direction when needed and translates it into medium-specific production controls. Produces prompt text but does not generate images. Do not use when a separate project-specific visual workflow owns the request or when the primary deliverable is broader content rather than an image prompt.