Claude Skills by alexeyshishin

Implement an already-approved plan from swarm-report/<slug>-plan.md. Routes each task to the executing agent that owns its file scope (frontend / backend / devops / mobile), runs the tests, writes an implementation report. Use AFTER /code-plan, never before.

Find the root cause of a bug or failing test. Spawns 1 debugger subagent that reproduces, ladders hypotheses, isolates with evidence, and applies the minimal fix. Use when a test fails, /code-build reports blocked, or "X stopped working". Does NOT design features (that's /code-plan).

Design a feature before writing code. Spawns planner + skeptic subagents that read the Memory Bank, argue, and write a plan to swarm-report/. Use BEFORE /code-build, for any feature touching 2+ files or with unclear scope. Skip for typos and one-line fixes.

Review a finished change against its plan. Spawns 1 reviewer subagent, reports ship/rework with severity-tagged findings. Use AFTER /code-build.

Turns an idea, a series of notes, or raw material into an article draft (Habr / Medium / personal blog) — with a TL;DR, a thesis → arguments → example → implication structure, and an authorial voice. Use when "write an article," "shape these thoughts into a long-read," "make a Habr post," "article draft."

Structures a technical tutorial — prerequisites, a "after reading this you'll be able to do X" goal, numbered steps with commands and checks, troubleshooting, cleanup. Doesn't write the full text: it produces a correct skeleton for the author to fill in with details. Use when "write a tutorial," "put together step-by-step instructions," "how-to," "tutorial," "deployment guide," "instructions for how to do X."

A content factory — turns a single source into a full content package in one pass. Takes as input a finished article, a video transcript, OR (if there's only a topic) offers to run deep research and assemble a source itself. Output: a 1500–2500 word article, 10 threads, 5 Reels/Shorts scripts, 5 posts (Telegram/VK/MAX), 3 carousels, and a content plan — in 6 separate files. Preserves the author's voice, avoids repetition between units (angle matrix), doesn't invent facts, and runs every text ...

A skill for humanizing Russian-language text. Removes signs of AI generation, makes the text feel alive. ALWAYS use this when the user asks to: humanize a text, remove traces of a neural network, make a text lively/natural, rewrite it like a human, humanize in Russian, remove officialese, remove wateriness/filler, make a text less formal. Also use this if the user pastes in Russian-language text and says something like 'rewrite this,' 'make it better,' 'sounds like a robot,' 'too artificial.'...

Turns an idea, a note, or raw text into a ready-to-post Telegram channel post — tight, thesis-first, in the author's voice, within 500-1500 characters. Use when the user wants "write a TG post," "shape this thought for the channel," "short post," "tg post," "shape an idea for subscribers."

Ultra-compressed communication mode. Cuts output tokens 65% (measured) by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.

Defragment and re-actualize a project memory bank (the .assistant/ knowledge base + the project's auto-memory). Reads everything that changed in the repo since the last defrag, finds where the memory bank has gone stale or accumulated amendment-on-amendment "patches", folds the patches into clean current-state docs, closes resolved questions, and updates the auto-memory — then shows a diff for review. Use when the user says "defrag the memory bank", "привести memory bank в порядок", "дефрагме...

Maintains a lightweight, always-current "project encyclopedia" in `.memory-bank/` — capturing architecture, stack, conventions, key entities, and active tasks. After each meaningful change, the AI suggests concise updates or explicitly confirms no update is needed.

Builds a Conventional Commits commit message for the current changes in git. Analyzes the staged diff (or unstaged, if staged is empty), determines type/scope, writes the subject and, if needed, a body. Use when the user wants to "make a commit", "commit the changes", "commit this", "write a commit message", "commit following the convention".

Builds a Pull Request description from the diff of the current branch against the base branch (main/master). Analyzes all commits, extracts "why / what / how to verify", captures breaking changes. Use when the user wants to "describe the PR", "write a description for the pull request", "put together a PR", "push and open a PR", "generate a PR description".

Enrich Obsidian note frontmatter — fill in the aliases, up, down, other fields without moving the note. Use this skill when the user asks to update/fill in a note's frontmatter, add aliases or links (up/down/other), or enrich a note with links — especially for notes in the `00. Входящие` folder. The note stays in place; only the frontmatter fields aliases, up, down, other change. Do not use for a full refactor with relocation — that's what obsidian-refactor-inbox is for.

Analyze and review notes from the inbox (`00. Входящие`) — inspection and reporting only, no file changes. The output is a list of notes with recommendations for category and action. For actually processing them (tags, links, moving) use obsidian-refactor-inbox.

Processes an arbitrary markdown file (article, web clip, transcript, notes, draft note) and turns it into structured knowledge for an Obsidian base using the PARA methodology. The skill extracts atomic entities (concepts, insights, facts), invents logical names for notes, creates a source literature note, and builds wikilinks and frontmatter. Use this skill whenever the user wants to "add this to the knowledge base", "rework this article", "process these notes", "extract the knowledge from th...

Critical analysis of Obsidian notes — hybrid search for similar and opposing topics in the knowledge base, identifying contradictions with existing knowledge, reasoned critique of argumentation and coverage completeness. Use this skill whenever the user asks to: critique a note, break it down, find contradictions, check the argumentation, find similar or alternative notes, do a peer review, \"what did I miss\", \"what's missing\", \"what's wrong\", \"find the gaps\". Also apply it after creat...

Process inbox notes in Obsidian. Use this skill when the user asks to process notes from the `00. Входящие` folder — updating tags, adding aliases, creating links to existing notes (people, MOCs, knowledge), moving them to the right folder per the PARA methodology. The skill determines the note type (atomic thought, resource, person, project, conference notes), adds the corresponding tags and links, and never deletes notes — only archives them with the #archive tag.

Refactor lecture notes in Obsidian — extract concepts, definitions, and theorems from lecture notes into separate notes, turn the lecture into a \"table of contents\" with links, update links to the discipline MOC. Use this skill when the user asks to process, split, or refactor a lecture's notes or several lectures. Triggers: \"process this lecture\", \"split these notes\", \"refactor the lecture\", \"pull the concepts out of this lecture\", \"process these notes\", \"split up the lecture\",...

Refactor a large Obsidian note — split it into atomic/thematic notes while preserving wikilinks in the original. Use this skill when the user asks to split, divide, or refactor a specific note. The skill creates child notes, inserts [[links]] into the original, and follows the PARA vault structure.

Offload an Obsidian hub note that too many notes link to (an in-degree tangle). The main signal of a tangle is a high in-degree (tens to hundreds of incoming links); out-degree is secondary. The skill creates sub-MOCs under the hub's existing categories and **redirects incoming links from resource notes** to the specific sub-MOC instead of the general one. The original MOC is kept as an entry point — it can still be linked to when the category is unclear. Use this skill whenever the user says...