agentleFS
olcayseyganolcayseygan/crumora/skills/know-me/SKILL.md

know-me

Skill0 starsChanged 43 days agoMITView on GitHub

Learns how you and this project work — repo, build, CI, git history, CLAUDE.md, memory and the prompts you typed into Claude Code here — and turns what repeats into the project's own Claude Code setup - skills, slash commands, subagents, hooks and CLAUDE.md lines. Every artifact cites its evidence; nothing for a need not yet seen, nothing duplicating what is installed, every hook run against a real payload. Output is one line per file with its evidence, a verification line, at most three watch-list lines and a restart note when skills or agents were written. Use when the user says "/know-me", "set up claude for this project", "generate skills for this repo", "make commands and hooks for this project", "what should I automate here", "build my .claude folder", "learn how I work", or the Turkish "beni tani", "projeyi tani", "bu proje icin skill hook command agent olustur", "claude kurulumunu yap". For checking code against a rule set use lint; for proving a request was understood use readback.

15 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use albert-dm, instantly-campaign and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • yawningwhaleAISkill

    albert-dm

    Reply to DMs and messages the way Albert would. Use when the user asks "how would Albert reply to this", pastes a conversation and wants an Albert-style response, or asks you to draft/write a DM in Albert's voice. Albert sells AI voice agents / automations to local service businesses (Go High Level, lead follow-up, bookings).

    Also in 5 projects that use this

  • yawningwhaleAISkill

    instantly-campaign

    Create a new Instantly cold-email campaign in the proven 5-email structure of the user's top-performing campaign (~16% reply rate). Use when the user wants to launch a new outbound campaign, write a cold email sequence, create an Instantly campaign, draft outreach for a new service or niche, or set up a sales sequence.

    Also in 6 projects that use this

  • RoxenRobloxSkill

    youtube-broll-maker

    Generate 10-second B-roll footage of yourself doing an action using the Higgsfield MCP (Seedance 2.0). Upload one photo of yourself, describe the action ("typing on laptop", "walking at night"), pick from 5 cinematography style presets, get an MP4. Use when the user wants B-roll, cutaway footage, "me doing X" video, YouTube B-roll, or any image-to-video of themselves performing an action.

    Also in 3 projects that use this

  • RoxenRobloxSkill

    youtube-popup-graphic

    Generate stunning Adobe-Premiere-Pro-style UI popup motion graphics for YouTube videos using the Higgsfield MCP. The user uploads a style reference image (any UI mockup, dashboard, glass card, design they love), names one skill/feature/item, and the skill produces an MP4 of a single card popping in styled to match the reference. Use when the user wants a "popup graphic", "popup animation", "card popup", "skill popup", "AE/Premiere-style animation", "stunning UI animation for YouTube", or says "make me a popup for <skill/feature/item> styled like <reference>".

    Also in 3 projects that use this

  • kronus-techSkill

    upwork-proposal

    Generate a tailored Upwork cover letter from a job posting. Invoke with the job posting text or URL. Auto-invoked when user says "cover letter", "upwork proposal", "apply to this job".

    Also in 5 projects that use this

  • op7418Skill

    youtube-clipper

    YouTube 视频智能剪辑工具。下载视频和字幕,AI 分析生成精细章节(几分钟级别), 用户选择片段后自动剪辑、翻译字幕为中英双语、烧录字幕到视频,并生成总结文案。 使用场景:当用户需要剪辑 YouTube 视频、生成短视频片段、制作双语字幕版本时。 关键词:视频剪辑、YouTube、字幕翻译、双语字幕、视频下载、clip video

    Also in 3 projects that use this

  • +4 more from yawningwhaleAI
  • +1 more from RoxenRoblox

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 3 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 14 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read how this project and this person work, find what repeats, and write the Claude Code setup that removes it: skills, slash commands, …

    line 8
  2. Distinctive:

    Evidence or nothing (MUST): Every artifact names the evidence that earned it — a file and line, a commit range, a transcript count, a …

    line 15
  3. Distinctive:

    Scope: Default: the session's repository, all five kinds.

    line 53
  4. Very specific:

    Verify (MUST): Every JSON and front matter block parses.

    line 75

“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

know-me — turn what repeats into the project's own setup

Read how this project and this person work, find what repeats, and write the Claude Code setup that removes it: skills, slash commands, subagents, hooks and CLAUDE.md lines. File formats, paths and hook payloads are in references/formats.md — read it before writing the first file.

Invariants

  • Evidence or nothing (MUST). Every artifact names the evidence that earned it — a file and line, a commit range, a transcript count, a CLAUDE.md rule. No generic starter-kit artifact, nothing for a workflow the project does not have yet.
  • Repetition is the bar. Three or more occurrences — in the prompts typed here, the git history, the CI steps — or a standing written rule in CLAUDE.md or memory. Twice goes on the watch list, not into a file.
  • Nothing that already exists (MUST). Inventory first: the project's CLAUDE.md files and .claude/ (settings, hooks, permissions, skills, commands, agents), the personal ~/.claude/ equivalents, installed plugins and their skills, the memory index, built-in commands. An overlap is dropped or folded in, never copied under a new name.
  • Reads are private, writes are project-scoped. Nothing from transcripts or memory is quoted into an artifact, except that trigger phrases in a description may use the user's own wording as typed. No secret, token, home-directory path or customer name lands in a generated file. Writes go to the project's .claude/ and CLAUDE.md; ~/.claude/ only when the user asks.
  • Existing files are merged, not replaced (MUST). settings.json is parsed, hooks added beside the existing ones, written back valid. CLAUDE.md gains lines; nothing the user wrote is reworded or removed. An existing skill, command or agent with the same name is left alone, the collision reported, and no new file takes that name.
  • Every hook is run before it is handed over (MUST). Real payload for its event, once with input that should pass and, where exit 2 blocks or feeds back, once with input that should trip it.
  • Generated code follows the project's own rules — commit conventions, language, style, lint.
  • One stop before writing. The inventory is confirmed with one yes before any file is written; the user may strike rows. No questions after the yes.
  • Lightest kind, fewest files. CLAUDE.md line → hook → command → skill → subagent; take the lightest that removes the need, and merge candidates that share a trigger.

Kinds

KindFitsNot when
CLAUDE.md linea fact or rule the model keeps getting wrong or being toldalready written, or the code states it
Hooka mechanical rule a script checks with no judgementthe check needs judgement
Slash commanda multi-line instruction the user starts on purposeit should trigger from plain speech
Skilla multi-step procedure that triggers from plain speechit fits in a paragraph
Subagenta recurring task that reads far more than it returns, or needs a narrower tool setit needs the conversation's context

Flow

  1. Scope. Default: the session's repository, all five kinds. An argument narrows it ("just hooks", a folder). Pinned once, never grows.
  2. Inventory what exists, one line each.
  3. Read the project — build surface, CI, layout, git history, written rules — only far enough to find what repeats.
  4. Read the person's prompts in this project's transcripts (path and entry shape in references/formats.md). Count by intent, across languages. Count corrections separately — a repeated correction is the strongest signal. Too few transcripts: say so in one line and work from the repo; never invent a habit.
  5. Propose and kill. Each candidate: kind, name, what it does, evidence, count. Drop any that breaks an invariant or can merge into another. Names in the project's language, no collision with a built-in command. Short only on count → watch list.
  6. The stop. One table, one question: write these?

    KindNameDoesEvidence
  7. Write each artifact in its format, with the project's real commands, paths and thresholds — no placeholders. Subagents get the narrowest tool list and an output contract, in the report style the project's CLAUDE.md sets, if it sets one. Hook scripts go in .claude/hooks/, use a runtime the project already has, read the payload from stdin, and exit 2 with a message naming the file and the rule when they block.
  8. Verify (MUST). Every JSON and front matter block parses; each name matches its folder or file. Every hook runs on a real payload for its event inside its timeout: pass case exits 0; where exit 2 blocks or feeds back (formats.md table), a trip case exits 2 with the expected message; on other events its output is checked. Every command and skill name is unique across project, personal and plugin scope. git diff --stat shows only the confirmed files. A red check is fixed and re-run.

Output

One line per file written, then the verification line, then at most three watch-list lines. Nothing else, except a restart note when skills or agents were written.

<path>    new|edit   <what it is>    [<evidence>]
verified: <JSON parsed>, <front matter parsed>, hook <name> pass=0 trip=2 | output ok, no name collisions, diff = <n> files
seen twice — worth watching: "<need>" (2x)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.