Stop naming other repositories in the docs and comments

.github/workflows/audit.yml

@@ -10,8 +10,7 @@
 # base image. The pipeline scans what a pull request builds, which catches a
 # change that introduces a vulnerable base. It cannot catch a base image tag
 # that stops receiving patches while nothing here changes, because nothing
-# runs. That is how a sibling repository came to publish 38 fixable high and
-# critical advisories.
+# runs, so an image can carry fixable high and critical advisories for weeks.
 name: audit
 
 on:

.github/workflows/pipeline.yml

@@ -208,9 +208,9 @@ jobs:
           tags: httphq:scan
 
       # The image half of `check:audit`. A base image pinned to a branch that
-      # has stopped receiving patches keeps building, and nothing says so: one
-      # repository in this fleet was publishing 38 fixable high and critical
-      # advisories that no change in it could have fixed.
+      # has stopped receiving patches keeps building, and nothing says so. The
+      # image then ships fixable high and critical advisories that no change
+      # here could fix.
       #
       # `os` packages only, because the application's own dependencies are
       # already gated by `check:audit`, and scanning everything reports the npm

AGENTS.md

@@ -218,20 +218,6 @@ and a whole-graph analysis over a subset reports almost everything as unused.
   closes the store. A new background task that writes to the store has to stop
   before the store closes; see `run` in `src/application.go`.
 
-## Shared files
-
-These files are kept identical, or in step, with the same files in the
-repositories that deploy to the lab (lab, 8ctave, geobear, postcraft,
-metric-tone and bjornkrols.com). Change them in all of them together.
-
-- `scripts/check-prose.mjs`, `scripts/check-lint-staged.mjs` and
-  `.husky/install.mjs`: identical in every one of them.
-- `src/logging`: in step with `apps/hooks/logging.go` in the lab repository,
-  which carries the same handler, redaction list and level rules.
-
-`knip.ts` is deliberately not shared, though every repository has one. It is
-almost entirely entry points, and no two repositories here have the same ones.
-
 ## Code layout
 
 `src` is a `main` package split by concern, one file per subject with its tests
@@ -303,12 +289,11 @@ columns, a `created_at`, and an index on every column a query filters or orders
 by. A `gorm:"column:..."` tag naming something else satisfies the struct and
 fails the test, which is the point.
 
-The sibling repositories run the same questions as SQL against Postgres after
-applying their migrations. There are no migration files here and no
-information_schema to query, so these go through `sqlite_schema` and the pragmas.
+There are no migration files here and no information_schema to query, so these
+questions go through `sqlite_schema` and the pragmas.
 
-Two of their checks do not apply. Nothing carries `updated_at`, because a
-capture is written once and never edited, and there is no email column anywhere.
+Nothing carries `updated_at`, because a capture is written once and never
+edited, so the test does not ask for one.
 
 ## Captured data is ephemeral
 

eslint.config.mjs

@@ -36,9 +36,7 @@ export default tseslint.config(
   // first of these.
   //
   // The modified variant scores a whole switch as one decision rather than one
-  // per case. It changes no number here, because none of these files switch,
-  // and it is stated so the rule reads the same as it does in the sibling
-  // repositories rather than differing in a way that looks deliberate.
+  // per case. It changes no number here, because none of these files switch.
   {
     rules: {
       complexity: ["error", { max: 4, variant: "modified" }],

scripts/check-lint-staged.mjs

@@ -8,7 +8,7 @@
 // cannot drift.
 //
 // Every statement fits in 80 columns, so Prettier prints this file the same at
-// any print width and the copies in the sibling repositories stay identical.
+// any print width.
 
 import { execFileSync } from "node:child_process";
 import { readFileSync } from "node:fs";

scripts/check-prose.mjs

@@ -9,9 +9,8 @@
 // `path` is a file, or a directory ending in "/". `pattern` is a regular
 // expression; its matches on that path are ignored.
 //
-// The same file lives in every repository that deploys to the lab. Change it
-// in all of them together. Every statement fits in 80 columns, so Prettier
-// prints it the same at any print width and the copies stay identical.
+// Every statement fits in 80 columns, so Prettier prints it the same at any
+// print width.
 import { execFileSync } from "node:child_process";
 import { existsSync, readFileSync } from "node:fs";
 

src/database/standards_test.go

@@ -10,13 +10,11 @@ import (
 )
 
 // The database standards, asked of the schema AutoMigrate actually produces
-// rather than of the struct tags that ask for it. The sibling repositories run
-// the equivalent as SQL against Postgres after applying their migrations; there
-// are no migration files here and no information_schema to query, so the same
-// questions are asked of sqlite_schema and the pragmas instead.
+// rather than of the struct tags that ask for it. There are no migration files
+// here and no information_schema to query, so the questions are asked of
+// sqlite_schema and the pragmas.
 //
-// Two of their four checks do not apply. There is no email column anywhere, and
-// nothing carries updated_at: a capture is written once and never edited, which
+// Nothing carries updated_at: a capture is written once and never edited, which
 // is the whole shape of this product, so a column nothing writes would be
 // overhead on the only table there is.