Skip to content
Back to search

all-city

SkillWorks with Claude CodeGitHub

Audit everything currently published under a name, as a set, and decide what stays. Use when someone asks what comes up when you google them or their company, when old pages are still live, when bios and taglines disagree across profiles, when a site is not showing up, when a rebrand or a name change leaves things behind, when someone is job hunting or fundraising and wants their presence to hold up, or when they say their online presence is a mess and they do not know where to start.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/michaeldoesmarketing--all-city/badge.svg)](/listings/michaeldoesmarketing--all-city)

HTML

<a href="/listings/michaeldoesmarketing--all-city"><img src="/api/listings/michaeldoesmarketing--all-city/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

git clone https://github.com/MichaelDoesMarketing/all-city.git ~/.claude/skills/all-city

Or grab the packaged `.skill` from [releases](https://github.com/MichaelDoesMarketing/all-city/releases).

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 154 lines
  1. ---
  2. name: all-city
  3. description: Audit everything currently published under a name, as a set, and decide what stays. Use when someone asks what comes up when you google them or their company, when old pages are still live, when bios and taglines disagree across profiles, when a site is not showing up, when a rebrand or a name change leaves things behind, when someone is job hunting or fundraising and wants their presence to hold up, or when they say their online presence is a mess and they do not know where to start.
  4. ---
  5. # All City
  6. Every skill that operates on one artifact assumes the artifact is what people will see. Usually it is not. People arrive through whatever they happen to land on, which is almost never the thing being maintained this week.
  7. This works on the whole standing body of work instead. Everything live right now that carries the name. Not the plan, not the next launch. What is up.
  8. Most people have never listed it. That is the job.
  9. Use the logic, do not perform the frame. Nothing in the output talks about graffiti, tags, or walls unless the user asks where the method comes from.
  10. ## The rule that protects this skill
  11. **Nothing goes on the list unverified.**
  12. Every surface in the inventory needs a URL that was actually opened and a note on what was found there. Not "you probably have a Crunchbase page." Not "there's likely an old About page." A confident inventory of surfaces that were never checked is worse than no inventory, because the user will act on it.
  13. Applied, that means:
  14. - A surface that cannot be opened is recorded as unverified, with the reason, and never given a verdict
  15. - A surface found through search is not verified until it has been opened
  16. - When a page cannot be reached, say which one and why, rather than describing what is probably on it
  17. - Never assert what a blocked page says. Blocked is the finding
  18. - Never recommend burying something that has not been looked at
  19. If most of the inventory could not be verified, say so plainly and deliver a short honest list instead of a long confident one.
  20. ## Step 1: Build the inventory
  21. Ask for what the user knows about, then go find the rest. People consistently forget the surfaces other people put up.
  22. Start with what they name, then work through `references/surfaces.md`, which lists the categories that get missed: syndicated copies, speaker and podcast pages, directory and data-broker listings, old subdomains, app store entries, archived campaign pages, co-authored posts, and anything a former employer still hosts.
  23. Then search the way a stranger would. Their name, their company, their name plus the category, the thing they want to be known for. Record what comes back and in what order. That order is data, and it is the input to Step 4.
  24. Mark each surface as **owned** (they can edit it), **claimed** (they control an account on someone else's platform) or **third-party** (they cannot change it). The verdict rules differ for each.
  25. ## Step 2: Check what each one actually does
  26. Run the script. It is in `scripts/surfacecheck.py`, needs nothing installed, and takes a file of URLs or a list of them.
  27. ```
  28. python3 scripts/surfacecheck.py urls.txt
  29. python3 scripts/surfacecheck.py https://example.com https://example.com/about
  30. python3 scripts/surfacecheck.py urls.txt --hand # self-descriptions only
  31. python3 scripts/surfacecheck.py urls.txt --json
  32. ```
  33. Each line of the file can be `Label | URL` to name the surface.
  34. It reports, per surface: whether it loads, where it actually lands after redirects, the title and description, how much text a crawler sees, whether the page needs JavaScript to say anything, and the newest year on the page. Then it lines up every surface's self-description in one block.
  35. That last part is the one that changes minds. Eleven different answers to "what do you do," side by side, is not an argument anyone wins.
  36. **Read the flags carefully.** `NO-RENDER` means a crawler got almost nothing, which usually means a client-rendered app. The page may look perfect in a browser and still be invisible to link previews, scrapers and some indexing. Confirm in a browser before telling someone their site is broken.
  37. If the script cannot run, open the surfaces by hand and record the same fields. Say that the numbers were gathered by hand.
  38. ## Step 3: Score the surface
  39. Four axes, 1 to 5. Score the body of work, not the individual pages.
  40. - **Coverage** - when someone looks for them, or for what they do, is there something of theirs there. Gaps matter more than quality
  41. - **Decay** - how much of what is up is stale, wrong, or contradicted by something newer
  42. - **Hand** - how consistently the surfaces describe the same person or company. A writer's tag is identical everywhere; most people's is not
  43. - **Order** - whether the things people hit first are the things they would choose
  44. Score Order last, because it depends on the search results from Step 1 rather than on any single page.
  45. `references/order.md` covers why first contact is rarely the homepage.
  46. ## Step 4: Give every surface a verdict
  47. Three options. No fourth, and no "monitor."
  48. - **KEEP** - it works, it is current, leave it alone. Say why, so the list is not only bad news
  49. - **FIX** - it earns its place but says the wrong thing, is stale, or is broken. Name the specific change
  50. - **BURY** - take it down, redirect it, or deindex it
  51. `references/verdicts.md` has the decision rules, including what to do about third-party surfaces that cannot be buried, where the move is to outrank rather than remove.
  52. **BURY is the one that matters.** People will rewrite a dead page forever rather than delete it. A surface with no traffic, no current claim, and no link value is costing maintenance and diluting everything else. Recommend removal and mean it.
  53. For anything third-party, never recommend contacting a publisher to remove a true, unflattering article. That is not what this is for. The move is to publish something better and let it rank.
  54. ## Step 5: Fix the hand before making anything new
  55. **No new work until the contradictions are resolved.**
  56. This is deliberately unwelcome, because making a new thing is more fun than fixing the bio on a profile nobody has opened in two years. It is still correct. A new piece landing on a surface that argues with itself is wasted, and the contradiction is what the reader remembers.
  57. Settle on one claim. Then push it to every owned and claimed surface in the same pass. `references/hand.md` covers what has to match and what is allowed to vary by platform.
  58. ## Modes
  59. **All city** - the default. Inventory, check, score, verdict, fix the hand.
  60. **First ten** - the fast uncomfortable one. Search the way a stranger would and report the first ten things they hit, in order, with nothing else. No recommendations. The order is the finding.
  61. **The buff** - decay only. Everything still up that is stale, wrong, or embarrassing. Use when someone is about to be looked up, before an interview, a launch, a raise, or a press cycle.
  62. **Hand check** - consistency only. Every surface's self-description in one block, then the one claim that should replace them.
  63. **Going up** - forward-looking. Something new is about to be published. Where will it land, what will it collide with, and does the existing surface support it or argue with it.
  64. ## Output format
  65. Deliver in this shape unless the user asks for something else.
  66. ```
  67. WHAT IS UP
  68. The inventory. Every surface, owned / claimed / third-party, with its URL
  69. and what was found. Unverified ones listed separately, with the reason.
  70. THE SCORE
  71. Coverage, Decay, Hand, Order. Four numbers, one line each.
  72. WHAT A STRANGER HITS FIRST
  73. The first few results in order, and whether that is the intended front door.
  74. HOW YOU DESCRIBE YOURSELF
  75. Every self-description side by side, verbatim.
  76. THE VERDICTS
  77. Each surface: KEEP, FIX or BURY, with the specific action.
  78. THE ONE CLAIM
  79. The single description that should replace the others, and where it goes.
  80. WHAT COULD NOT BE CHECKED
  81. Everything unverified, and why. Never empty unless everything was opened.
  82. ```
  83. Worksheets are in `assets/surface-sheet.md` and `assets/first-ten.md`.
  84. ## Rules of engagement
  85. **Verified or it does not exist.** See the rule above. It is the only one here with no exceptions.
  86. **Old is not the same as stale.** A piece from four years ago that still makes a true claim and still gets traffic is an asset. Age alone is never a reason to bury.
  87. **Presence is not performance.** Being on a platform is not a win if nobody looks for them there. A thin profile on a network they do not use is dead weight, not coverage.
  88. **Do not audit what they did not ask about.** Personal accounts, private profiles, and anything a reasonable person would consider separate from the work stay out unless the user puts them in scope.
  89. **Never recommend suppressing true information.** The response to an accurate unflattering surface is better work ranking above it, never removal.
  90. **Third-party surfaces get a different verdict set.** They cannot be fixed or buried directly, only outranked, corrected through the platform's own process, or accepted. Say which.
  91. **Say what is working.** An output that is only a list of problems gives a false picture and will not get run twice. KEEP verdicts are part of the deliverable.
README.md · README · 102 lines
  1. # all-city
  2. **Everything you ever published is still up.**
  3. You remember the thing you made this week. You don't remember the conference bio you sent in 2023, the landing page from a campaign that ended, the profile on a platform you quit, or the article someone else wrote about you that outranks your own site.
  4. Strangers don't arrive at your front door. They arrive wherever the link they clicked went, and that link is almost never the page you maintain. Whatever they land on is the entire thing they know about you.
  5. `all-city` takes an inventory of everything currently live under your name, scores it as a set, and makes you decide what stays.
  6. ## How it works
  7. Four axes, scored on the body of work instead of any single page.
  8. **Coverage.** When someone searches for you, or for the thing you want to be known for, is there anything of yours there. Gaps count for more than quality.
  9. **Decay.** What's still up but stale, wrong, or quietly contradicted by something newer. Decay isn't age. A piece from five years ago that still makes a true claim is an asset.
  10. **Hand.** A graffiti writer's tag is identical on every wall. Thousands of reps, same letterforms, recognizable from a moving car. Your bio probably isn't. Most people have a different answer to "what do you do" on every surface they own, written at a different time, never compared.
  11. **Order.** What people actually hit first, versus what you'd have chosen. These are rarely the same and almost nobody has looked.
  12. ## It actually checks
  13. There's a Python script in the folder. No dependencies, takes a file of URLs or a list of them.
  14. Per surface it reports whether it loads, where it really lands after redirects, the title and description, how much text a crawler sees, whether the page says anything at all without JavaScript, and the newest year on the page.
  15. Then it stacks every surface's self-description in one block. Eleven different answers to "what do you do," side by side, is not an argument anyone wins.
  16. ## The verdicts
  17. Every surface gets one of three. **KEEP**, **FIX**, or **BURY**. There's no fourth, and specifically no "monitor," which is how a broken page survives another two years while appearing on a list.
  18. BURY is the one that does the work. The instinct is always to rewrite rather than remove, so a dead page gets a fresh coat every eighteen months and keeps helping nobody.
  19. Then the rule people don't want: **nothing new gets made until the contradictions are resolved.** Fixing a bio nobody has opened in two years is less fun than making a new thing. It's still correct. A new piece landing on a surface that argues with itself inherits the argument.
  20. ## The rule that keeps it honest
  21. **Nothing goes on the list unverified.**
  22. Every surface needs a URL that was actually opened. Not "you probably have a Crunchbase page." Not "there's likely an old About page somewhere." A confident inventory of surfaces nobody checked is worse than no inventory, because you'll act on it.
  23. Anything that can't be opened goes in a separate section with the reason, and never gets a verdict. Blocked is a finding, not a gap to fill with a guess.
  24. It also won't help you suppress something true. The answer to an accurate unflattering article is better work ranking above it. Never a takedown.
  25. ## Install
  26. ```bash
  27. git clone https://github.com/MichaelDoesMarketing/all-city.git ~/.claude/skills/all-city
  28. ```
  29. Or grab the packaged `.skill` from [releases](https://github.com/MichaelDoesMarketing/all-city/releases).
  30. ## Try it with
  31. > "What comes up when someone googles me?"
  32. > "I'm interviewing next week. Is anything embarrassing still live?"
  33. > "Does my LinkedIn say the same thing as my site?"
  34. Five modes. **All city** is the full pass. **First ten** searches you the way a stranger would and lists what they hit, in order, with no advice attached. **The buff** is decay only, for when you're about to be looked up on purpose. **Hand check** stacks your self-descriptions. **Going up** runs before you publish something new, to see what it's about to land on top of.
  35. ## Real demo: I ran it on myself
  36. I have a personal site, a blog on it, a LinkedIn I post to, an X account, and eight public repos. I'd have told you my presence was in decent shape.
  37. **The headline finding.** My homepage and my blog return byte-identical HTML. Same 14,057 bytes, same title, same meta description, same everything. A crawler asking for either one gets **97 characters** of text. My browser renders **10,355**.
  38. So the long-form writing that is the entire point of the site doesn't exist to anything that doesn't run JavaScript. Every blog URL presents as the homepage. Link previews, scrapers, and anything indexing without rendering see one page where I think there are many.
  39. **The one I'd never have found.** My own site's Twitter card credits someone else:
  40. ```
  41. twitter:site @Lovable
  42. twitter:image https://michael-does-marketing.lovable.app/og-image.png
  43. og:url https://www.michaeldoesmarketing.com/ (page served at the bare domain)
  44. ```
  45. Anyone who shares my site on X credits the tool I built it with. The preview image is hosted on a different domain than the site. The canonical points at `www` while the page answers without it. Default template values nobody edited, including me, for however long it's been live.
  46. **Three surfaces I can't audit at all.** My LinkedIn profile, a LinkedIn article, and my X profile are all disallowed to automated fetching. Not a script problem. That's the actual condition: a large share of my presence is invisible to anything that isn't a logged-in human, which means it can't be previewed, indexed normally, or checked. They're listed as unverified with the reason, and they get no verdict. That's the rule working.
  47. **The scores.**
  48. ```
  49. Coverage 3/5 Present on the big surfaces. Searching the position I've spent
  50. months building returns nothing of mine.
  51. Decay 4/5 Little that's stale. Most of it is recent.
  52. Hand 3/5 Three related but different claims across site, LinkedIn and
  53. GitHub. Same theme, three different sentences.
  54. Order 3/5 The name finds me. The thing I want to be known for doesn't.
  55. ```
  56. Decent-looking presence. The front door is invisible to machines and credits a third party, and I'd never have checked, because the site looks perfect in my browser. That's the whole argument for this one.
  57. ## License
  58. MIT. Use it, fork it, ship it.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.