All authors
Claude Skills by Arthuro0103
github.com/Arthuro010317 skills0 installs15 views
- Cf Adapt SkillGuides the user to make a skill of THEIR OWN from a catalog worksheet (vault/skill-worksheets/) - read the worksheet, separate the method from what belongs only to the original author, decide whether it solves one of their pains, write it with their answers, test it on a real case and record the verdict in vault/memory/skills-reviewed.md. One question per message. Claude never writes the skill alone, and "does not fit me" is a valid answer that gets recorded. TRIGGERS - "/cf-adapt-skill", "qu...Votes: 0GitHub stars: 10
- Cf AdoptApplies the cogiforge to an Obsidian vault the person ALREADY has, in place, copying nothing into an inbox. Runs tools/adopt.py as a dry run, reads the plan in plain words (areas inferred from their folders, what would be installed, the debt measured today, and that the hook only judges notes in the commit), asks "want to apply?", and only then runs it with --apply --yes. It invents no link and states nothing about the person. TRIGGERS - "/cf-adopt <vault>", "use this on my existing vault", "...Votes: 0GitHub stars: 10
- Cf AskAnswers a question about the user's own workbench using ONLY what the notes say, and cites every claim with [[path#Heading]] that a script then checks exists. Rewrites the question into 3-5 short queries, retrieves passages with tools/ask.py (local BM25, no API, no network), discards the irrelevant ones saying why, answers from what is left. If no passage supports the answer it says "the vault does not cover this". TRIGGERS - "ask the vault", "what do my notes say about X", "what do I know ab...Votes: 0GitHub stars: 10
- Cf BrainstormGeneral brainstorming for any product, project, study, event or loose idea (pains are one possible input, not the only one). Asks about the goal one question at a time, scans what the vault already says (collection before internet) and declares the gap, proposes 3 genuinely different paths, compares them in a table by criteria the person picks (defaults - value, frequency of the problem, effort, dependence on others, reversibility), recommends one WITHOUT choosing for them, and closes in a br...Votes: 0GitHub stars: 10
- Cf Claude CornerClaude's corner. When the user says they are leaving, Claude uses the time away (whatever they say, at most 1h30) to reread the notes, find real connections, read the pains and the projects, and test ideas in a throwaway place outside vault/. Only proposes, in vault/memory/corner/YYYY-MM-DD-<slug>.md - it does not assert, does not merge, does not push, does not install anything outside the throwaway, does not send messages to anyone. ORDER RULE - if the user asked for something before leaving...Votes: 0GitHub stars: 10
- Cf Close SessionCloses the workday on the workbench. Writes the day's diary (vault/memory/diary/), one briefing per session (vault/memory/briefings/) with a full-path link to the project root in the body, collects pains from the user's own words into vault/memory/ideas/_pains.md, opens a task in vault/tasks/ for whatever is left pending, and runs the verifiers if they exist. A new skill or new project comes out only as a proposal. Commits only if the user says so and never pushes. TRIGGERS - "fecha a sessão"...Votes: 0GitHub stars: 10
- Cf Extract RoutinePulls a real routine out of the user's head by interview and writes it down so someone else can run it. One question at a time, context before options, a short answer is valid. Covers expected result, trigger, inputs, steps in real order, decisions, exceptions, who does it and proof of done. Unknowns are marked [CONFIRM]; the agent never invents a step. States - draft, ready-for-human-review (the most the agent may declare) and validated (only after a SECOND person follows it without guessing...Votes: 0GitHub stars: 10
- Cf Import KnowledgeBrings an existing knowledge base (another vault, a folder of notes, a wiki export, loose documents) into the workbench. Runs tools/import.py, reads the report it writes, and then helps triage the imported inbox into vault/notes/ one batch at a time, following "the target comes first": a named target and a real link in the body for every note that moves. It proposes, the user approves; it invents no link and states nothing about the user. TRIGGERS - "/cf-import-knowledge <folder>", "import my...Votes: 0GitHub stars: 10
- Cf Know My ProductInterviews the user about their product, one question at a time, and fills the product brief of ONE project in vault/projects/<name>/product.md using only their own words (who uses it, the problem, what exists, what it is NOT, success metric, constraints). Collects pains voiced during the interview with the literal quote and the date, proposes a ranking by frequency and cost, and links each pain to a project or task. Claude proposes; nothing is written to memory before the user says "yes". TR...Votes: 0GitHub stars: 10
- Cf New ProjectCreates a new project in the user's workbench without hand-copying the example. Asks one question at a time (the name in a few words, then "what exists in the world when this project succeeds?", then type and status with defaults), copies the user's sentence word for word into declared_target, shows a --dry-run, waits for an explicit "yes", then runs tools/new_project.py, which writes the project root, adds the row to projects/_index.md, runs hub.py, and then gate and ring are run and their r...Votes: 0GitHub stars: 10
- Cf OnboardWorkbench onboarding. Asks questions that follow what the user answered, one at a time, and the output comes from their own words - fills vault/_questions_about_me.md and proposes lines for vault/memory/profile.md with a date and a verbatim quote. Deduces nothing the user did not say; a short answer is a valid answer, and so is "prefer not to say". TRIGGERS - "/cf-onboard", "primeira vez aqui", "atualiza meu perfil", "get to know me", "how do I use this workbench". May be suggested (never imp...Votes: 0GitHub stars: 10
- Cf Open SessionOpens a work session on one workbench project. Takes a project NAME, reads its root (vault/projects/<name>/instructions.md), the latest diary entries and its open tasks, and returns in a few lines where the project stands and what the last session left behind. Reports only what is in the files; when the file's status disagrees with what the disk shows, it becomes a question, never a verdict. TRIGGERS - "abrir sessão do <projeto>", "/cf-open-session <project>", "onde parei no <projeto>", "resu...Votes: 0GitHub stars: 10
- Cf StudyHelps a student learn what they want, the way they want. Interviews one question at a time (what, why, deadline, HOW they like to learn, time per day), asks three level questions, writes projects/<topic>/study-plan.md, then loops: explain a small block, ask an active-recall question, log each item in projects/<topic>/mastery.md with the date and the student's own words as proof. An item is learning until it passes a RETEST on a DIFFERENT day; only then it is mastered. Never claims the student...Votes: 0GitHub stars: 10
- Cf Task ObserverObserves a work session and ONLY PROPOSES process improvements, appending to vault/memory/observations.md (date, what was seen, what is proposed). Looks for three signals - a skill that was missing, a step repeated three times, a correction the user made twice. Never writes to profile.md, patterns.md, decisions.md nor vault/notes/, and never creates a skill: the skill agenda belongs to the user, and cf-adapt-skill guides the creation. It is the "improves with me" half of the workbench. TRIGGE...Votes: 0GitHub stars: 10
- Cf Whats RealKeeps vault/projects/<name>/whats-real.md, a page with three sections - Works (each item with proof - the command run, its output and the date), Simulated or demo (what looks like it works and does not) and Unknown - plus a short dated diary.md for the project. It is created BEFORE a milestone (demo, delivery, class, launch) and CHECKED after it - every Works item needs its proof re-run, otherwise it goes back to Unknown. The agent never moves an item to Works without proof. TRIGGERS - "/cf-w...Votes: 0GitHub stars: 10
- Skill WorksheetsA skill is a **folder with a `SKILL.md` file** in `.claude/skills/<name>/`. Claude Code reads the `description` of **every** skill at the start of each session and only opens the body of the one that matches the request. So two things matter a lot: the `description` decides **when** the skill triggers, and each line of it **costs context in every session**, even in the ones where it never triggers. To see it in practice, open `.claude/skills/cf-onboard/SKILL.md` alongside (the shortest one) a...Votes: 0GitHub stars: 10
- Cf BrandInterviews the user, one question at a time, and writes the brand pieces of a person or a project using only their own words - voice rules for the person (vault/memory/voice.md) or for one project (inherits the person's and keeps only what differs), and for a project the DNA, the messaging, the design system (design.md plus tokens.json, with no default color) and a manual that assembles them. Verbatim quotes, nothing deduced, a "yes" before anything is saved. Checks texts and previews tokens ...Votes: 0GitHub stars: 10