Registry audit

What audit_everything() does

audit_everything() walks the replication registry and attempts every table and figure in each listed engine (R and Stata where both exist). It is meant as a health check: some failures are expected as studies, data, and dependencies change over time.

Key behaviour:

• patience (default 20 seconds; registry reports often use 60) — each table or figure is halted after this audit cap; the audit continues with the next object. Timeout rows record timeout_seconds and an explicit “Timed out after N seconds (audit cap)” message.
• Failures do not stop the run — results are collected in a data frame.
• Incomplete / unavailable steps are skipped — yaml incomplete: true (including requires_engine: / data_unavailable: gaps) is not attempted. Rows are recorded with status Skipped and a reason, and are not counted as success or failure. Distinct from fail/timeout.
• Report fields — study, object id, engine, success, skipped flag, elapsed seconds, timeout_seconds (audit cap), timed-out flag, and a short error / skip-reason snippet.
• Substantive checks (default substantive = TRUE) — when a study defines tests/substantive/<step_id>.R, the audit compares replicated estimates to published benchmarks (see Fearon & Laitin tab_1). Failures appear as [substantive] in the printed summary.
• Related studies are compiled when the registry index is rebuilt ([refresh_registry()] / internal index build): upstream from paper.related / paper.extends, downstream by reversing those pointers. Use summary(get_study(doi)) for a quick console view; the audit itself does not re-fetch related links per run.

library(replicateEverything)

# Point at a local monorepo checkout (optional)
options(
  replicateEverything.registry_root = "/path/to/replicate_everything/registry",
  replicateEverything.study_folders_root = "/path/to/replicate_everything",
  replicateEverything.use_sibling_packages = TRUE
)

audit <- audit_everything(patience = 20)
print(audit)

Incremental CSV job store

As of replicateEverything 0.7.40+, registry audits no longer replace the whole portfolio snapshot on every call. When a registry root is set, each run upserts into a flat CSV and rebuilds derived files from the full store.

Source of truth: registry/audit_jobs.csv

File	Role
audit_jobs.csv	Source of truth — one row per job, keyed by doi × object × engine
audit_summary.json	Derived portfolio counts + progress (Shiny health bar)
audit_latest.rds	Derived full audit_everything object rebuilt from the CSV

Upsert means: rows for jobs in this run are updated; every other study’s rows stay put. A one-DOI or one-collection audit therefore does not wipe the Shiny health bar or drop other studies from the derived summary.

Maintainer verbs (0.7.41)

Four public maintainer verbs cover registration and audit:

Verb	Role
[register_study()]	Validate + sync one study stub into the registry
[refresh_registry()]	Light by default: rebuild index + Shiny cache, seed CSV gaps, rebuild summary. Pass audit = TRUE or audit = <dois> for live runs
[audit_everything()]	Manual / heavy / Quarto live audit
[audit_report()]	Read-only portfolio health from CSV / summary (no writes)

# Light refresh (no live engines)
refresh_registry("registry")

# Light refresh + full live audit
refresh_registry("registry", audit = TRUE, patience = 20)

# Light refresh + subset live audit
refresh_registry("registry", audit = "10.1257/aer.91.5.1369")

# Read-only console / dataframe view
audit_report("registry")

Full portfolio vs add / refresh a subset

# Full registry (every index row)
audit <- audit_everything(patience = 20)

# Add or refresh only these DOIs / collections (upsert; others kept)
audit <- audit_everything(patience = 30, dois = "10.1257/aer.91.5.1369")
audit <- audit_everything(patience = 20, collections = "APSR")

After either kind of run, disk holds the full portfolio: the run merges into audit_jobs.csv, then rebuilds JSON and RDS from every CSV row.

Seed without live runs

[refresh_registry()] (default audit = FALSE) seeds gaps so the health bar is not empty before the first full audit. It does not execute replications: prior audit_latest.rds, bake timings / artifact mtimes fill missing rows; existing source = "audit" rows are kept.

A normal audit_everything() already upserts and refreshes derived files; you do not need a separate rebuild after a live audit.

last_success_at

Each CSV row carries last_success_at:

1. On a successful audit for that job → set to this run’s finish time.
2. Otherwise → keep the prior CSV value when present.
3. If still empty → fall back to bake timing recorded_at, else the local artifact’s mtime.

Failed or timed-out passes therefore do not erase the last known success time.

How to read audit outputs

Source	What it represents
audit_jobs.csv	Full portfolio job history (canonical). Inspect or edit here.
audit_summary.json	Baked counts + progress buckets for the Shiny health bar. Prefer this for first paint / lightweight consumers.
audit_latest.rds	Full audit_everything object derived from the entire CSV (not just the last call’s subset).
In-memory return of audit_everything()	This run only — if you passed dois = / collections =, $results and $summary cover that subset. Do not treat the return value as the portfolio when writing vignette snapshots or reports.

For vignette or package snapshots, copy the registry RDS (full portfolio), not saveRDS(audit) from a filtered call:

file.copy(
  file.path(getOption("replicateEverything.registry_root"), "audit_latest.rds"),
  "inst/vignette-data/audit_latest.rds",
  overwrite = TRUE
)

Quarto report

The Quarto document in the registry repo still works the same way. Render it (or scripts/run_audit.R) to run a live audit and write the HTML report. After write, disk always holds the full portfolio CSV + derived JSON/RDS, even when params restrict dois / collections.

# Quarto report lives in the registry repo (sibling registry/ in a monorepo)
quarto::quarto_render(audit_everything_qmd(), execute_params = list(patience = 20))
quarto::quarto_render(
  audit_everything_qmd(),
  execute_params = list(patience = 20, collections = "APSR")
)

# Or: live R audit, then HTML from saved RDS (no second live run)
# Rscript scripts/run_audit.R
# quarto render audit_everything.qmd -P refresh:false

Console progress tags (0.7.39)

With verbose = TRUE (default), each finished job prints a short status tag aligned with the health-bar buckets:

Tag	Meaning
ok	Runnable job succeeded (including substantive checks when defined)
timeout	Hit the patience audit cap
substantive_fail	Replication ran but published-value check failed
missing_engine	Engine / dependency gap (e.g. Stata not found)
other	Other failure or skip bucket

Example line: - Table 1 (table, r) [ok].

Commit checklist (registry repo)

After an audit (or seed + refresh) that you want on GitHub / Shiny:

1. audit_jobs.csv — always commit with the derived files
2. audit_summary.json — health bar
3. audit_latest.rds — full snapshot
4. audit_everything.html (+ audit_everything_files/ if present) — only if you are publishing the Quarto report

Push the registry repository so remote consumers see the new summary. Optional: copy audit_latest.rds into replicateEverything/inst/vignette-data/ and rebuild pkgdown so this article’s tables stay current.

Latest audit results

The table below is built from a saved audit snapshot shipped with the package (inst/vignette-data/audit_latest.rds). Set REPLICATE_AUDIT_LIVE=true (and point at a local monorepo) only when you intend to refresh that snapshot.

audit <- if (run_live) {
  tryCatch({
    monorepo <- Sys.getenv("REPLICATE_MONOREPO", unset = "")
    if (!nzchar(monorepo)) {
      parent <- normalizePath(
        file.path(find.package("replicateEverything"), "..", ".."),
        mustWork = FALSE
      )
      if (file.exists(file.path(parent, "registry", "index.csv"))) {
        monorepo <- parent
      }
    }
    if (nzchar(monorepo)) {
      options(
        replicateEverything.registry_root = file.path(monorepo, "registry"),
        replicateEverything.study_folders_root = monorepo,
        replicateEverything.use_sibling_packages = TRUE
      )
    }
    audit_everything(patience = 20, verbose = FALSE)
  }, error = function(e) {
    message("Live audit skipped: ", conditionMessage(e))
    NULL
  })
} else {
  NULL
}

if (is.null(audit) && nzchar(audit_rds) && file.exists(audit_rds)) {
  audit <- readRDS(audit_rds)
}
if (is.null(audit)) {
  stop("No audit results available.")
}
sm <- audit$summary
results <- audit$results

Summary

Metric	Value
Patience (seconds per object)	60
Studies audited	14
Replication runs	96
Successful	85
Failed	10
Timed out	7
Skipped	1
Audit started	2026-07-29 10:53
Audit finished	2026-07-29 11:10

attempted <- sm$runs - (sm$skipped %||% 0L)
if (attempted > 0) {
  pct <- round(100 * sm$success / attempted, 1)
  cat(sprintf("**Pass rate (excluding skipped):** %s%%\n", pct))
}
#> **Pass rate (excluding skipped):** 89.5%

Results by study

if (!"skipped" %in% names(results)) {
  results$skipped <- FALSE
}
studies <- unique(results$title)
for (study in studies) {
  cat("\n\n#### ", study, "\n\n", sep = "")
  sub <- results[results$title == study, , drop = FALSE]
  sub$status <- replicateEverything:::audit_result_status(
    sub$success, sub$timed_out, sub$skipped
  )
  sub$seconds <- ifelse(is.na(sub$seconds), NA, round(sub$seconds, 2))
  show <- sub[, c(
    "object_label", "object", "engine", "status", "seconds", "error_snippet"
  )]
  names(show) <- c("Object", "ID", "Engine", "Status", "Seconds", "Error")
  print(knitr::kable(show, row.names = FALSE))
}

A Welfare Analysis of Policies Impacting Climate Change

Object	ID	Engine	Status	Seconds	Error
Figure 1	fig_1	stata	Timed out	60.11	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
Figure 2	fig_2	stata	Timed out	60.12	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
Figure 3	fig_3	stata	Timed out	60.11	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
Figure 5	fig_5	stata	Timed out	60.21	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
Figure 6	fig_6	stata	Timed out	60.11	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
Figure 4	fig_4	stata	Failed	3.96	Expected Stata output not found. | Expected file: C:/WZB Dropbox/Macartan Humphreys/5_github/replicate_everything/rep-10.1257-aer.20250166/outputs/mvpf_subsidies_Fig4_scc193.png | Stata ran: yes | Executable: C:/Program Files/Stata17/StataM…
Figure 7	fig_7	stata	Failed	3.56	Expected Stata output not found. | Expected file: C:/WZB Dropbox/Macartan Humphreys/5_github/replicate_everything/rep-10.1257-aer.20250166/outputs/mvpf_taxes_Fig7_scc193_with_CIs.png | Stata ran: yes | Executable: C:/Program Files/Stata17/S…
Figure 8	fig_8	stata	Failed	3.58	Expected Stata output not found. | Expected file: C:/WZB Dropbox/Macartan Humphreys/5_github/replicate_everything/rep-10.1257-aer.20250166/outputs/mvpf_intl_Fig8_scc193_with_CIs.png | Stata ran: yes | Executable: C:/Program Files/Stata17/St…
Table 1	tab_1	stata	Timed out	60.22	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
Table 2	tab_2	stata	Timed out	60.30	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit

Beyond Belief Change: The Persuasive Returns of Targeting Attitude-Relevant Beliefs

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	r	OK	11.96
Table 3	tab_3	r	OK	3.99

Bounding Causes of Effects With Mediators

Object	ID	Engine	Status	Seconds	Error
Figure 1	fig_1	r	OK	0.22

COVID-19 vaccine acceptance and hesitancy in low- and middle-income countries

Object	ID	Engine	Status	Seconds	Error
Table 2	tab_2	r	OK	0.33
Fig. 1	fig_1	r	OK	5.05
Fig. 2	fig_2	r	OK	3.81
Fig. 3	fig_3	r	OK	7.83
Extended Data Fig. 1	ext_fig_1	r	OK	8.52
Extended Data Fig. 2	ext_fig_2	r	OK	41.73

Ethnicity, Insurgency, and Civil War

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	r	OK	0.53
Table 1	tab_1_stata	stata	OK	5.40

Ideological Alignment and Evidence-Based Policy Adoption

Object	ID	Engine	Status	Seconds	Error
Table H.1	tab_h1	stata	Skipped	NA	Table H.1 not available because of proprietary data
Table 1	tab_1	stata	OK	9.42
Table 2	tab_2	stata	OK	5.52
Table 3	tab_3	stata	OK	5.25
Table 4	tab_4	stata	OK	5.27
Figure 2	fig_2	stata	OK	19.39

Migration, Families, and Counterfactual Families

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	stata	OK	21.06
Table 2	tab_2	stata	OK	13.47
Table 3	tab_3	stata	OK	11.32
Table A.1	tab_A_1	stata	OK	22.15
Table A.2	tab_A_2	stata	OK	19.72
Table A.3	tab_A_3	stata	OK	19.94
Table A.4	tab_A_4	stata	OK	20.40
Table A.5	tab_A_5	stata	OK	19.40
Figure 1	fig_1	stata	OK	5.42
Figure 1	fig_1_panel_b	stata	OK	5.96
Figure A.2	fig_A_2	stata	OK	5.76
Figure 2	fig_2_panel_a	stata	OK	6.36
Figure 2	fig_2_panel_b	stata	OK	6.47
Figure A.1	fig_A_1_panel_a	stata	OK	7.22
Figure A.1	fig_A_1_panel_b	stata	OK	7.45
Figure 3	fig_3_panel_a	stata	OK	4.88
Figure 3	fig_3_panel_b	stata	OK	5.02
Figure 4	fig_4_panel_a	stata	OK	4.94
Figure 4	fig_4_panel_b	stata	OK	5.11

Minimal folder-backed template study

Object	ID	Engine	Status	Seconds	Error
Simple table	tab_1	r	OK	0.1

Minimal reanalysis template repo

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	r	OK	0.22

Portraits of Power: Facial Appearance and the Tacit Domain of Political Selection in China

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	stata	OK	8.43
Table 2	tab_2	stata	OK	7.28
Table 3	tab_3	stata	OK	7.12
Figure 2	fig_2	python	OK	30.74
Figure 4	fig_4	r	OK	4.45
Figure 5	fig_5	r	OK	3.51

Preventing Rebel Resurgence after Civil War: A Field Experiment in Security and Justice Provision in Rural Colombia

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	stata	OK	4.97

Public support for global vaccine sharing in the COVID-19 pandemic: Evidence from Germany

Object	ID	Engine	Status	Seconds	Error
Figure 1	fig_1	r	OK	4.50
Figure 2	fig_2	r	OK	8.41
Figure 3	fig_3	r	OK	1.22
Figure 4	fig_4	r	OK	1.31
Figure 5	fig_5	r	OK	1.88
Figure 6	fig_6	r	OK	2.61
Figure 7	fig_7	r	OK	4.47
Figure 8	fig_8	r	OK	4.67
Table 2	tab_2	r	OK	4.00
Table 1	tab_1	r	OK	3.70

Sovereignty, Substance, and Public Support for European Courts’ Human Rights Rulings

Object	ID	Engine	Status	Seconds	Error
Table 3a — Summary statistics, Deportation vignette	tab_3a	r	OK	1.42
Table 3b — Summary statistics, Quran burning and Eviction vignettes	tab_3b	r	OK	1.20
Figure 1 — Treatment summaries	fig_1	r	OK	9.62
Figure 2a — Effect of EC disagreeing with domestic court (H1)	fig_2a	r	OK	3.00
Figure 2b — Effect of EC disagreeing with domestic court, by country	fig_2b	r	OK	3.93
Figure 2c — Effect of EC disagreeing with domestic court, by country (averaged over vignettes)	fig_2c	r	OK	2.64
Figure 3a — H1 heterogeneity by domestic rule-of-law satisfaction	fig_3a	r	OK	6.47
Figure 3b — H1 heterogeneity by domestic rule-of-law satisfaction (UK/Denmark only)	fig_3b	r	OK	4.48
Figure 3c — Difference in EC-disagreement effect by rule-of-law satisfaction (UK/Denmark only)	fig_3c	r	OK	3.03
Figure 4a — Effect of case outcome (H2)	fig_4a	r	OK	3.31
Figure 4b — Difference between case outcome effect and sovereignty effect	fig_4b	r	OK	2.73
Figure 5a — H2 interactions by sympathy toward applicant	fig_5a	r	OK	6.97
Figure 5b — Difference in outcome effect between sympathetic and unsympathetic respondents	fig_5b	r	OK	4.44
Figure 6 — Heterogeneity by nationalism	fig_6	r	OK	11.63
Figure 7 — Heterogeneity by authoritarianism	fig_7	r	OK	14.08

The Colonial Origins of Comparative Development

Object	ID	Engine	Status	Seconds	Error
Table 1	tab_1	r	OK	0.77
Table 1	tab_1_stata	stata	OK	1.85
Table 2	tab_2	r	OK	0.16
Table 2	tab_2_stata	stata	OK	2.60
Table 3	tab_3	r	OK	0.15
Table 3	tab_3_stata	stata	OK	2.86
Table 4	tab_4	r	OK	0.23
Table 4	tab_4_stata	stata	OK	2.98
Table 5	tab_5	r	OK	0.23
Table 5	tab_5_stata	stata	OK	3.12
Table 6	tab_6	r	OK	0.19
Table 6	tab_6_stata	stata	OK	3.60
Table 7	tab_7	r	OK	0.20
Table 7	tab_7_stata	stata	OK	3.53
Table 8	tab_8	r	OK	0.31
Table 8	tab_8_stata	stata	OK	3.96

Failures (concise)

if (!"skipped" %in% names(results)) {
  results$skipped <- FALSE
}
fails <- results[results$success %in% FALSE & !results$skipped %in% TRUE, , drop = FALSE]
if (nrow(fails) == 0) {
  cat("All recorded runnable jobs succeeded (or were skipped).\n")
} else {
  fails$seconds <- ifelse(is.na(fails$seconds), NA, round(fails$seconds, 2))
  fails$status <- replicateEverything:::audit_result_status(
    fails$success, fails$timed_out, fails$skipped
  )
  show <- fails[, c(
    "title", "object_label", "object", "engine",
    "status", "seconds", "error_snippet"
  )]
  names(show) <- c(
    "Study", "Object", "ID", "Engine", "Status", "Seconds", "Error"
  )
  knitr::kable(show, row.names = FALSE)
}

Study	Object	ID	Engine	Status	Seconds	Error
A Welfare Analysis of Policies Impacting Climate Change	Figure 1	fig_1	stata	Timed out	60.11	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
A Welfare Analysis of Policies Impacting Climate Change	Figure 2	fig_2	stata	Timed out	60.12	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
A Welfare Analysis of Policies Impacting Climate Change	Figure 3	fig_3	stata	Timed out	60.11	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
A Welfare Analysis of Policies Impacting Climate Change	Figure 5	fig_5	stata	Timed out	60.21	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
A Welfare Analysis of Policies Impacting Climate Change	Figure 6	fig_6	stata	Timed out	60.11	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
A Welfare Analysis of Policies Impacting Climate Change	Figure 4	fig_4	stata	Failed	3.96	Expected Stata output not found. | Expected file: C:/WZB Dropbox/Macartan Humphreys/5_github/replicate_everything/rep-10.1257-aer.20250166/outputs/mvpf_subsidies_Fig4_scc193.png | Stata ran: yes | Executable: C:/Program Files/Stata17/StataM…
A Welfare Analysis of Policies Impacting Climate Change	Figure 7	fig_7	stata	Failed	3.56	Expected Stata output not found. | Expected file: C:/WZB Dropbox/Macartan Humphreys/5_github/replicate_everything/rep-10.1257-aer.20250166/outputs/mvpf_taxes_Fig7_scc193_with_CIs.png | Stata ran: yes | Executable: C:/Program Files/Stata17/S…
A Welfare Analysis of Policies Impacting Climate Change	Figure 8	fig_8	stata	Failed	3.58	Expected Stata output not found. | Expected file: C:/WZB Dropbox/Macartan Humphreys/5_github/replicate_everything/rep-10.1257-aer.20250166/outputs/mvpf_intl_Fig8_scc193_with_CIs.png | Stata ran: yes | Executable: C:/Program Files/Stata17/St…
A Welfare Analysis of Policies Impacting Climate Change	Table 1	tab_1	stata	Timed out	60.22	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit
A Welfare Analysis of Policies Impacting Climate Change	Table 2	tab_2	stata	Timed out	60.30	Timed out after 60 seconds (audit cap; timeout_seconds: 60) | ! Native call to processx_wait failed | Caused by error in chain_clean_call(...): | ! reached elapsed time limit

Skipped (incomplete / unavailable)

if (!"skipped" %in% names(results)) {
  results$skipped <- FALSE
}
skips <- results[results$skipped %in% TRUE, , drop = FALSE]
if (nrow(skips) == 0) {
  cat("No steps were skipped as incomplete / unavailable.\n")
} else {
  skips$status <- "Skipped"
  show <- skips[, c(
    "title", "object_label", "object", "engine", "status", "error_snippet"
  )]
  names(show) <- c(
    "Study", "Object", "ID", "Engine", "Status", "Reason"
  )
  knitr::kable(show, row.names = FALSE)
}

Study	Object	ID	Engine	Status	Reason
Ideological Alignment and Evidence-Based Policy Adoption	Table H.1	tab_h1	stata	Skipped	Table H.1 not available because of proprietary data

Interpreting failures

Common reasons a run fails or times out:

• Missing study package or folder — install the study repo locally or set replicateEverything.study_folders_root to your monorepo root.
• Stata not installed — Stata-backed entries fail until Stata is found; see the Stata replications vignette.
• Network / data — folder-backed studies may need data files downloaded on first run.
• Patience / audit cap — slow jobs may time out under the configured patience (e.g. 60s in the registry report). That is recorded as Timed out with timeout_seconds and an explicit audit-cap message, not as a silent failure.
• Skipped steps — Mathematica / proprietary / incomplete yaml steps appear as Skipped, not Failed or Timed out.

Re-run locally and refresh the package vignette snapshot:

Sys.setenv(REPLICATE_AUDIT_LIVE = "true")
options(
  replicateEverything.registry_root = "/path/to/replicate_everything/registry",
  replicateEverything.study_folders_root = "/path/to/replicate_everything",
  replicateEverything.use_sibling_packages = TRUE
)
audit <- audit_everything(patience = 20)
# Prefer the registry portfolio RDS (derived from audit_jobs.csv), not a
# filtered in-memory return value:
file.copy(
  file.path(getOption("replicateEverything.registry_root"), "audit_latest.rds"),
  "inst/vignette-data/audit_latest.rds",
  overwrite = TRUE
)

# Full HTML report: quarto render audit_everything.qmd (in the registry repo)

After adding or updating registry stubs, maintainers typically run [refresh_registry()] so index.csv, the Shiny cache, and the audit summary stay in sync:

refresh_registry("../registry")                      # light
refresh_registry("../registry", audit = TRUE, patience = 20)
audit_report("../registry")