zebrabi/zebra-bi-agentic / plugins/zebra-bi/skills/report-authoring
report-authoring
VerifiedPlan, author and verify Zebra BI reports (Tables+, Charts+, Cards, certified Tables and Charts) in PBIP/PBIR from the terminal: frame the reader and the comparison (analyst layer), choose visual form, size, scale and sort (IBCS design layer), write the TMDL model, then bindings, variance semantics, formatting, small multiples, comment layers, pre-flight and the screenshot verify loop. Use when creating or editing a Power BI report with Zebra BI visuals, building from a saved .pbip or a flat spreadsheet, or debugging a Zebra visual that renders blank, black, wrong-numbered or ignores a property. Also answers which comparison, period, fiscal year or polarity a page needs, and what a P&L, cost-centre, sales, pipeline, cash, headcount or SaaS page should carry. Triggers: Zebra BI, zebraBiCards, ZebraBITables, ZebraBITablesPlus, ZebraBIChartsPlus, waterfall6C9ED82, Tables+, Charts+, Plus visuals, viewer commenting, IBCS report, variance chart, small multiples, publicCustomVisuals, PBIP, PBIR, TMDL, Desktop bridge.
| File | Kind | Size |
|---|---|---|
| .bundle-sha256 | other | 1.0 KB |
| SKILL.md | skill md | 75.1 KB |
| references/analyst.md | markdown | 39.9 KB |
| references/dax.md | markdown | 13.4 KB |
| references/design.md | markdown | 51.0 KB |
| references/formatting.md | markdown | 29.2 KB |
| references/model-layer.md | markdown | 48.2 KB |
| references/preflight.md | markdown | 20.9 KB |
| references/recipes.md | markdown | 26.8 KB |
| references/tmdl.md | markdown | 48.1 KB |
| references/verify-loop.md | markdown | 42.2 KB |
| references/visuals.md | markdown | 46.5 KB |
| rules.json | config | 29.6 KB |
references/preflight.md
View on GitHub<!-- Summary: Run this after every write and before every launch — the mechanical PBIR and TMDL faults that render silently wrong or block a cold open, caught in about a second instead of a 90s restart or a human spotting them in a screenshot. -->
# Pre-flight — catch it at authoring time, not in the screenshot
Every check here exists because the fault it catches **actually shipped** and was found
by a human looking at a render. That is the expensive way. All of them are mechanical.
**Run after every write. Run before every launch.** It costs ~1 second. A cold-open failure costs
~90s, and a silent corruption costs a review cycle.
This is the **whole-project** check: report layer *and* model layer. It supersedes the model-only
lint in `model-layer.md`, which is the same TMDL checks without the report half — run this one
and you do not need that one. Why each TMDL check exists in full (what the error message looks like,
how to fix it) is still documented there. The script carries the eight TMDL killers that cost a cold
open most often plus four of the cheap grammar checks (a duplicate object, a `///` separated from its
object by a blank line, a relationship endpoint naming nothing in the model, a `database.tmdl` that does
not start with its `database` line); `validate_report` carries those twelve plus the remaining grammar
rules (unknown property, illegal enum value, duplicate property, and the rest of `references/tmdl.md`
section 7), so when the tool is available it is the fuller check.
## Why a *report-layer* check is needed at all
The skill's oldest rule is *"the screenshot is the only evidence"* — true, and it stays true. But it
has been read as *"only the screenshot can find problems"*, which is wrong and expensive. A whole
class of fault is **mechanically detectable before you ever launch Desktop**: corrupted text, a BOM,
unparseable JSON, a missing page file. Leaving those to the eye wastes the render loop on things a
regex settles, and — worse — they are the faults most likely to be *missed* by eye, because a
mangled title still looks like a title.
## The script
```powershell
param([string]$Project) # folder containing <name>.pbip
$issues = @()
# ---------- THE .pbip SHORTCUT ----------
# Desktop validates this BEFORE it looks at the model or the report, and rejects a wrong $schema by
# opening `Untitled` with an empty canvas plus a Frown dialog - a third cause of a silent `Untitled` open, and the only one a regex settles. The natural wrong guess is the shape
# every OTHER PBIR file uses (`item/report/definition/...`); the .pbip does not follow it.
foreach ($f in (Get-ChildItem $Project -Filter *.pbip -File)) {
try {
$s = ([System.IO.File]::ReadAllText($f.FullName, [Text.Encoding]::UTF8) | ConvertFrom-Json).'$schema'
} catch { $issues += "BAD JSON: $($f.Name)"; continue }
if ($s -notmatch '^https://developer\.microsoft\.com/json-schemas/fabric/pbip/pbipProperties/1\.\d+\.\d+/schema\.json$') {
$issues += "PBIP SCHEMA: $($f.Name) has '$s' - Desktop opens Untitled. Expected .../fabric/pbip/pbipProperties/1.0.0/schema.json"
}
}
# ---------- REPORT LAYER ----------
$rep = Get-ChildItem $Project -Directory -Filter *.Report | Select-Object -First 1
if ($rep) {
foreach ($f in (Get-ChildItem $rep.FullName -Recurse -Filter *.json)) {
$bytes = [System.IO.File]::ReadAllBytes($f.FullName)
# 1. BOM - PBIR rejects it, and reload names the file but only once you launch
if ($bytes.Length -ge 3 -and $bytes[0] -eq 239 -and $bytes[1] -eq 187 -and $bytes[2] -eq 191) {
$issues += "BOM: $($f.FullName)"
}
# 2. MOJIBAKE - the one that shipped. Text decoded as ANSI and re-encoded as UTF-8,
# usually by a ConvertFrom/ConvertTo-Json round-trip that did not pin the encoding.
# A mangled title still renders and still looks like a title.
# MATCH ON BYTES, NOT LITERALS: 0xC3 0x83 is the double-encoding signature. Pasting the
# mojibake characters into a .ps1 breaks the PowerShell parser - learned the hard way.
$hi = 0; $mojibake = $false
for ($i = 0; $i -lt $bytes.Length; $i++) {
if ($bytes[$i] -gt 127) { $hi++ }
if ($i -lt $bytes.Length - 1 -and $bytes[$i] -eq 0xC3 -and $bytes[$i+1] -eq 0x83) { $mojibake = $true }
}
if ($mojibake) { $issues += "MOJIBAKE (double-encoded): $($f.FullName)" }
elseif ($hi -gt 0) { $issues += "NON-ASCII ($hi bytes) - keep PBIR titles ASCII: $($f.FullName)" }
# 3. Parses at all
$text = [System.Text.Encoding]::UTF8.GetString($bytes)
try { $null = $text | ConvertFrom-Json } catch { $issues += "BAD JSON: $($f.FullName)" }
}
# 3b. TITLE TOKEN GATE - a [Measure] token in titleSettings.text resolves ONLY if that
# measure is projected into the visual's Filters role. Otherwise it renders the
# literal "[LFL label]" on the page and looks like a typo nobody made.
foreach ($vf in (Get-ChildItem $rep.FullName -Recurse -Filter visual.json)) {
$v = [System.IO.File]::ReadAllText($vf.FullName, [Text.Encoding]::UTF8) | ConvertFrom-Json
# reset per iteration and guard the null array - textboxes have no titleSettings, and a
# leaked $title from the previous visual produces a confident false positive
$title = $null
$ts = $v.visual.objects.titleSettings
if ($ts -and $ts.Count -gt 0) { $title = $ts[0].properties.text.expr.Literal.Value }
if (-not $title) { continue }
$inFilters = @()
if ($v.visual.query.queryState.Filters) { $inFilters = @($v.visual.query.queryState.Filters.projections.nativeQueryRef) }
foreach ($m in [regex]::Matches($title, '\[([^\]]+)\]')) {
$tok = $m.Groups[1].Value
if ($inFilters -notcontains $tok) {
$issues += "TOKEN NOT RESOLVABLE: '[$tok]' in $($vf.Directory.Name) title; Filters role has [$($inFilters -join ', ')]"
}
}
}
# 4. Every page in pageOrder has a page.json, and activePageName is one of them
$pagesFile = Join-Path $rep.FullName 'definition\pages\pages.json'
if (Test-Path $pagesFile) {
$pages = [System.IO.File]::ReadAllText($pagesFile, [Text.Encoding]::UTF8) | ConvertFrom-Json
foreach ($pg in $pages.pageOrder) {
if (-not (Test-Path (Join-Path $rep.FullName "definition\pages\$pg\page.json"))) {
$issues += "MISSING PAGE: $pg is in pageOrder with no page.json"
}
}
if ($pages.activePageName -and $pages.pageOrder -notcontains $pages.activePageName) {
$issues += "activePageName '$($pages.activePageName)' is not in pageOrder"
}
}
# 5. A base theme or a custom theme (Desktop writes only a customTheme on Save As PBIP and that opens fine), or the report renders NOTHING with no error
$reportJson = Join-Path $rep.FullName 'definition\report.json'
if (Test-Path $reportJson) {
$rj = [System.IO.File]::ReadAllText($reportJson, [Text.Encoding]::UTF8) | ConvertFrom-Json
if (-not $rj.themeCollection.baseTheme -and -not $rj.themeCollection.customTheme.name) { $issues += "NO THEME - the report will render blank with no page tabs" }
foreach ($pkg in $rj.resourcePackages) {
# SharedResources are Desktop built-ins (e.g. baseTheme CY23SU08) - no file on disk is correct.
# Verified by render: a report declaring one with no file paints normally.
if ($pkg.type -eq 'SharedResources') { continue }
foreach ($item in $pkg.items) {
if (-not (Test-Path (Join-Path $rep.FullName "StaticResources\$($pkg.name)\$($item.path)"))) {
$issues += "MISSING RESOURCE: $($item.path) is declared but absent"
}
}
}
}
}
# ---------- MODEL LAYER ----------
$sm = Get-ChildItem $Project -Directory -Filter *.SemanticModel | Select-Object -First 1
if ($sm) {
$measures = @{}; $columns = @{}; $objs = @{}; $ends = @(); $allCols = @{}
foreach ($f in (Get-ChildItem "$($sm.FullName)\definition" -Recurse -Filter *.tmdl)) {
$b = [System.IO.File]::ReadAllBytes($f.FullName)
if ($b.Length -ge 3 -and $b[0] -eq 239 -and $b[1] -eq 187 -and $b[2] -eq 191) { $issues += "BOM: $($f.Name)" }
$lines = [System.IO.File]::ReadAllLines($f.FullName); $tbl = ""
$inFence = $false; $exprIndent = -1
for ($i = 0; $i -lt $lines.Count; $i++) {
$ln = $lines[$i]; $t = $ln.Trim()
$ind = (($ln -replace "`t", ' ') -replace '\S[\s\S]*$', '').Length
# `//` IS legal DAX, so only a `//` outside an expression body is a TMDL fault. A fenced
# body OPENS at the END of the assignment line and closes on a line that is only the fence.
if ($inFence) { if ($t -eq '```') { $inFence = $false; $exprIndent = -1 }; continue }
if ($t.EndsWith('```') -and $t -ne '```') { $inFence = $true; continue }
$inExpr = $false
if ($exprIndent -ge 0) {
if ($t -eq '') { $inExpr = $true } elseif ($ind -gt $exprIndent) { $inExpr = $true } else { $exprIndent = -1 }
}
if (-not $inExpr -and $ln -match "^\s*(?:[A-Za-z_]\w*\s+)?(?:'[^']*'|[A-Za-z_][\w .\-]*)\s*=") { $exprIndent = $ind }
# These three are guarded by the FENCE only, not by $inExpr, and the difference matters.
# $inExpr marks every deeper line after any assignment as expression body -- right for `//`,
# and it would make a property-position check unable to fire under `measure X = 1`, which is
# exactly where `description:` appears. A fence is the only place a TMDL file holds arbitrary
# text. Screened against 1,325 real .tmdl files: zero hits with no guard at all.
if ($t -match '^description\s*:') {
$issues += "DESCRIPTION PROPERTY: $($f.Name):$($i+1) - TMDL has no 'description:' property; use a /// line above the object. Fails with 'Unsupported property - description is not a supported property in the current context!'" }
if ($t -match "^(table|column|measure|partition|hierarchy|level|role|perspective|calculationGroup|calculationItem)\s+(.+)$") {
$kw = $Matches[1]; $nm = $Matches[2]
$eq = $nm.IndexOf('=')
if ($eq -ge 0) { $nm = $nm.Substring(0, $eq) }
$nm = $nm.Trim()
# Only the NAME. Everything after `=` is DAX and its spaces are ordinary.
if ($nm -and -not $nm.StartsWith("'") -and $nm -match '\s') {
$issues += "UNQUOTED NAME: $($f.Name):$($i+1) - $kw $nm - a name containing a space must be quoted ('$nm'). Desktop reports this as an INDENTATION error on a correctly indented line" } }
if ($inExpr) { continue }
if ($t -match '^//(?!/)') { $issues += "COMMENT: $($f.Name):$($i+1) - '//' is not TMDL" }
if ($t -match '^///') {
$gap = $false
for ($j = $i + 1; $j -lt $lines.Count; $j++) {
$n = $lines[$j].Trim()
if ($n -eq '') { $gap = $true; continue }
if ($n -match '^///') { break }
if ($n -match '^relationship\b') {
$issues += "DESC ON RELATIONSHIP: $($f.Name):$($i+1) - relationships take no /// description; DataModelLoadFailed" }
# A /// description sits DIRECTLY above its object. A blank line between them fails the load.
elseif ($gap) { $issues += "DESCRIPTION GAP: $($f.Name):$($i+1) - a blank line separates this /// description from its object" }
break
}
}
if ($ln -match '^table\s+(.+)$') { $tbl = $Matches[1].Trim().Trim("'")
if ($tbl -eq 'Measures') { $issues += "RESERVED NAME: $($f.Name) - 'Measures' is a reserved table name (ModelSchemaValidationFailed); use '.Measures'" }
# Unlike 'Measures' this one OPENS. The model is fine and every visual bound to the table
# renders "Something's wrong with one or more fields", so nothing connects the two.
if ($tbl -eq 'Goal') { $issues += "RESERVED NAME: $($f.Name) - 'Goal' loads without complaint and then breaks every visual bound to it; rename it (Goals, Target)" } }
# A multi-line expression must not start on the `=` line: the `=` is bare and the body is
# indented under it. Two forms continue below -- unbalanced parens, and VAR (which needs a
# RETURN, so it always does). Parens are counted outside string literals so a caption cannot
# open a phantom continuation, and ``` is skipped because that IS the legal multi-line form.
if ($t -match "^measure\s+(?:'([^']+)'|([^\s=]+))\s*=\s*(\S.*)$") {
$mn = if ($Matches[1]) { $Matches[1] } else { $Matches[2] }
$expr = $Matches[3].Trim()
if (-not $expr.StartsWith('```') -and $i + 1 -lt $lines.Count) {
$nx = $lines[$i + 1]
if ($nx.Trim() -ne '') {
$nxInd = (($nx -replace "`t", ' ') -replace '\S[\s\S]*$', '').Length
if ($nxInd -gt $ind) {
$depth = 0; $q = $false
foreach ($c in $expr.ToCharArray()) {
if ($c -eq '"') { $q = -not $q }
elseif (-not $q) { if ($c -eq '(') { $depth++ } elseif ($c -eq ')') { $depth-- } } }
if ($depth -gt 0 -or $expr -match '^VAR\s') {
$issues += "MULTILINE MEASURE: $($f.Name):$($i+1) - measure '$mn' starts DAX on the '=' line and continues below. The '=' must be bare, body indented deeper. Fails with 'Unexpected line type: Other!'" } } } } }
if ($t -match "^measure\s+'?([^'=]+?)'?\s*=") { $measures["$tbl|$($Matches[1].Trim())"] = 1 }
if ($t -match "^column\s+'?([^'\r\n]+?)'?\s*$") { $columns["$tbl|$($Matches[1].Trim())"] = 1 }
# A measure, column, hierarchy or partition is declared ONCE per table, across every file of the
# model. The second declaration fails the load; a name typed twice in a split model is how it happens.
if ($t -match "^(measure|column|hierarchy|partition)\s+(?:'([^']+)'|([^\s=]+))") {
$on = if ($Matches[2]) { $Matches[2] } else { $Matches[3] }
# Every column, calculated ones (`column X = <DAX>`) included, is a legal relationship end.
if ($Matches[1] -eq 'column') { $allCols["$tbl|$on"] = 1 }
$ok = "$tbl|$($Matches[1])|$on"
if ($objs.ContainsKey($ok)) { $issues += "DUPLICATE OBJECT: $($f.Name):$($i+1) - $($Matches[1]) '$on' is declared twice in table '$tbl' (first at $($objs[$ok]))" }
else { $objs[$ok] = "$($f.Name):$($objs.Count)" } }
# Relationship endpoints, collected here and resolved after every file is read.
if ($t -match "^(fromColumn|toColumn):\s*(.+)$") { $ends += ,@("$($f.Name):$($i+1)", $Matches[2].Trim()) }
}
}
foreach ($k in $measures.Keys) { if ($columns.ContainsKey($k)) { $issues += "NAME COLLISION: $k" } }
# Both ends of a relationship must name a table and a column the model declares, as Table.Column
# (the table quoted when its name has a space). An endpoint naming nothing fails the whole load.
foreach ($e in $ends) {
$ref = $e[1]
if ($ref -match "^'([^']+)'\.(.+)$") { $rt = $Matches[1]; $rc = $Matches[2].Trim().Trim("'") }
elseif ($ref -match "^([^.]+)\.(.+)$") { $rt = $Matches[1]; $rc = $Matches[2].Trim().Trim("'") }
else { $issues += "RELATIONSHIP ENDPOINT: $($e[0]) - '$ref' is not Table.Column"; continue }
if (-not $allCols.ContainsKey("$rt|$rc")) { $issues += "RELATIONSHIP ENDPOINT: $($e[0]) - '$ref' names no column the model declares" }
}
# database.tmdl begins with its `database` line, named or bare. A file holding only properties fails.
$dbf = Join-Path $sm.FullName 'definition\database.tmdl'
if (Test-Path $dbf) {
$first = (Get-Content $dbf | Where-Object { $_.Trim() -ne '' } | Select-Object -First 1)
if ($first -notmatch '^database(\s|$)') { $issues += "DATABASE LINE: database.tmdl must start with 'database' (found '$first')" }
}
# Reported, not flagged: a cultures/ file is legitimate TMDL that cold-opens perfectly and, on
# Desktop 2.156, silently killed `file.reload/v1`, i.e. the whole one-second loop (one 2.159
# observation reloaded fine). Knowing costs nothing; not knowing cost seven cold opens.
$cul = Join-Path $sm.FullName 'definition\cultures'
if (Test-Path $cul) {
"NOTE: a cultures/ file is present - on Desktop 2.156 file.reload/v1 did not work on such a project (it cold-opens normally; one 2.159 observation reloaded fine). Try a report-only reload, and if it fails budget a cold open per iteration or remove the file if nothing needs it."
}
# Reported, not flagged. Every shipped template sits at 1600/1601 and every one of them opens,
# so this is not a defect — it only means `file.reload/v1` is refused until the file says 1606.
$db = Join-Path $sm.FullName 'definition\database.tmdl'
if (Test-Path $db) {
$m = [regex]::Match((Get-Content $db -Raw), 'compatibilityLevel:\s*(\d+)')
if ($m.Success -and [int]$m.Groups[1].Value -lt 1606) {
"NOTE: compatibilityLevel $($m.Groups[1].Value) - Desktop runs the model at 1606 or higher, so a reload of a file below 1606 is refused as a downgrade. Write 1606 (up to 1702 is accepted). Not a fault."
}
}
}
if ($issues.Count -eq 0) { "PRE-FLIGHT: PASS" } else { "PRE-FLIGHT: $($issues.Count) issue(s)"; $issues | % { " - $_" } }
```
## Writing files without creating the faults in the first place
- **Do not hand-author ANY schema-versioned file.** `.pbip`, `definition.pbir`, `version.json`,
`definition.pbism` and `.platform` are three lines each, they carry a pinned `$schema`, and there is
no upside to writing one from memory. **Copy a working project's copy and change the path/name**
(and re-stamp `.platform`'s `logicalId` so the copy has its own identity). Guessing a `$schema` by
analogy with the neighbouring files is exactly how the check above came to exist.
- **Never `Set-Content -Encoding utf8`** in PowerShell 5.1 — it writes a BOM. Use
`[System.IO.File]::WriteAllText($p, $t, (New-Object System.Text.UTF8Encoding($false)))`.
- **Assert the shape of GENERATED data before you write it.** Building DAX/JSON rows in a loop and
joining them is where silent corruption enters: a multi-line PowerShell `@(...)` array literal fed
to `-join ','` emitted 700 **space**-separated `DATATABLE` rows — valid-looking,
completely wrong, and it would have surfaced as an unexplained cold-open failure. Use a
`System.Collections.Generic.List[string]` + `[string]::Join`, then **assert one row** (count the
separators) before writing. Same family as the `ConvertTo-Json` blob damage below.
- **Never round-trip a `visual.json` through `Get-Content | ConvertFrom-Json | ConvertTo-Json`
without pinning the encoding on the way in.** `Get-Content` guesses, and each guess-and-rewrite
compounds: a `·` became `·`, then `ÃÆ'Æ’…` after several passes. Read with
`[System.IO.File]::ReadAllText($p, [Text.Encoding]::UTF8)`.
- **Better: keep report-layer titles ASCII.** A `-` separator survives every toolchain; `·` and `—`
are one careless read away from mojibake and the render still *looks* fine.
## Three name-resolution checks worth more than everything else here
Each of these is a **cold-open failure or a visible error in the report**, each is decided entirely
by text you already have on disk, and each costs a full restart to find by rendering. All three fire
on the same mistake: **renaming something and not updating what points at it.** That is not an
exotic error — it is the single most common edit an author makes.
**1. Every relationship endpoint must resolve.** For each `relationship` in `relationships.tmdl`,
split `fromColumn` and `toColumn` on the dot and confirm the table has a file and the column is
declared in it. A stale endpoint gives:
> Cannot resolve all the paths while de-serializing Database. Property FromColumn of object
> "relationship X" refers to an object which cannot be found
…and the project does not open at all.
**2. Every report field reference must resolve.** Walk every `visual.json` for `Measure` and
`Column` nodes, read `Expression.SourceRef.Entity` and `Property`, and confirm that table exists and
declares that measure or column. This one does **not** stop the report opening — it renders, and the
field is silently absent, with `Missing_References` in the error detail. **A report that opens is
not a report whose fields resolved.**
**3. A multi-line measure must put NOTHING after the `=`.** TMDL takes the body on the following
lines, indented deeper than the property:
```tmdl
measure 'AR overdue' =
VAR AsAt = [AR as at]
RETURN
COUNTROWS ( FILTER ( 'AR', 'AR'[Settled] = 0 && 'AR'[DueDate] < AsAt ) )
formatString: #,0
```
Start the DAX on the `=` line and continue it below and **the whole model fails to load**, with no
message naming the measure. Detect it as: a `measure` line with text after `=`, whose next line is
indented three tabs or more and is not `formatString` / `lineageTag` / `displayFolder` / `isHidden`.
## Mechanical, but NOT checked here yet
Known-detectable faults this script does not cover. Do them by hand until someone writes and
**renders** a check for them:
- **Textbox height vs font size.** `h >= fontSize(pt) * 1.333 * 1.3 + ~16` — a 22pt title in a 40px
box gets Power BI's own scrollbar and clips.
- **Visuals overlapping, or hanging off a 1280×720 canvas.** Renders without complaint.
## What this does NOT replace
The screenshot. Everything Zebra-specific — bindings, roles, whether a number is right, whether the
variance is comparable — is invisible to this script and always will be. This catches the
*mechanical* layer so the render loop is spent on the *semantic* one.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |