panoma 0.1.9 → 0.9.0

package/dist/chunk-FN63LYZH.js

@@ -1,825 +0,0 @@
-// src/messages.ts
-var MESSAGES = {
-  "server.unreachable": "Couldn\u2019t reach the catalog at {api}.",
-  "server.startIt": "Start it with: panoma up",
-  "server.badApi": "\u201C{api}\u201D is not a valid address.",
-  "server.notThisMachine": "{host} is not this machine.",
-  "server.remoteHint": "`panoma up` only starts the local catalog; a remote one is started by whoever hosts it.",
-  "server.alreadyUp": "The catalog was already up \xB7 {api}",
-  "server.noMonorepo": "Can\u2019t find Panoma\u2019s monorepo from here.",
-  "server.noMonorepoHint": "The web server lives in the source, not in the CLI package ({entry}).\nStart it from your clone: pnpm --filter @panoma/web run dev",
-  "server.starting": "Starting the catalog\u2026",
-  "server.startingDev": "Starting the web server\u2026 the first run compiles, so it may take a while.",
-  "server.noPnpm": "Couldn\u2019t launch pnpm. Is it installed?",
-  "server.pnpmFailed": "Couldn\u2019t launch pnpm: {reason}",
-  "server.pnpmHint": "Panoma starts the web app with pnpm, so it has to be on the PATH.",
-  "server.timedOut": "The server didn\u2019t answer within 60 seconds.",
-  "server.oldDatabase": "Your catalog was written by an older panoma (PostgreSQL {written}; this one uses {current}).",
-  "server.oldDatabaseHint": "The format changes between major Postgres versions. Your data is intact in {path}:\ndelete nothing. The conversion lives in the repository, at ops/migrate-pglite5.mjs.",
-  "server.portBusy": "The port for {api} is already taken by something else.",
-  "server.portBusyHint": "Stop whatever is there, or start this one elsewhere: panoma up --api http://localhost:4174",
-  "server.childDied": "The server exited on its own (code {code}). The last thing it wrote:",
-  "server.staleVersion": "A panoma {running} is running and the installed one is {installed}.",
-  "server.staleVersionHint": "Updating does not replace the live process. Restart it: panoma down && panoma up",
-  "server.otherInstance": "There is already a panoma at {api} (pid {pid}) using this same database.",
-  "server.otherInstanceHint": "Two at once corrupt it. Stop that one with panoma down, or give this one its own PANOMA_HOME.",
-  "server.databaseBusy": "Another process ({who}) already has this catalog's database open.",
-  "server.databaseBusyHint": "Two at once corrupt it, whoever started them. Stop that one first, or give this one its own PANOMA_HOME. To see it yourself: lsof +D {db}",
-  "server.databaseLeasedHint": "Two at once corrupt it, whoever started them. Stop that one first, or give this one its own PANOMA_HOME. It left its mark at {lease}",
-  "server.timedOutHint": "What it said while starting up is in {log}",
-  "server.up": "Catalog up \xB7 {api}",
-  "server.upDetail": "pid {pid} \xB7 log at {log}",
-  "server.upStop": "stop it with: panoma down",
-  "server.networkOpen": "Open to your local network, with a key.",
-  "server.networkLicence": "Sharing it over the network is what makes the AGPL ask you to offer the source to whoever opens it.\nThe catalog already links to it in the sidebar; if you modified it, publish your changes too.",
-  "server.networkHint": "Anyone with these links is in your catalog. Open one once and that browser stays in:",
-  "server.networkHere": "this machine",
-  "server.networkThere": "your network",
-  "server.networkLocalToo": "While the port is open, localhost needs the key too: from the network anyone can claim to be localhost.",
-  "server.networkOperator": "Only the first link can run anything here \u2014 install, build, open an editor. The network link is for looking.",
-  "server.none": "No server was started by panoma up.",
-  "server.somethingElse": "But something answers at",
-  "server.somethingElseHint": "something else started it, and I won\u2019t touch that.",
-  "server.gone": "Process {pid} is gone. Forgetting the pid and touching nothing else.",
-  "server.stopped": "Catalog stopped (pid {pid})",
-  "server.bootUnsupported": "--on-boot isn\u2019t written for {platform} yet.",
-  "server.bootUnsupportedHint": "It is written for macOS (a LaunchAgent), Linux (a systemd user unit) and Windows\n(a task at logon). Not for the others, and pretending otherwise would be worse\nthan saying so.",
-  "server.bootNeedsBuild": "Starting at login needs the CLI to be built.",
-  "server.bootNeedsBuildHint": "Build it with: pnpm --filter panoma build",
-  "server.bootInstalled": "At login, the catalog starts itself.",
-  "server.bootNotLoaded": "it was written, but could not be activated. By hand: {command}",
-  "server.bootLog": "boot output goes to",
-  "server.bootRemove": "to remove it:",
-  "today.badReport": "The catalog replied with something that isn\u2019t a report.",
-  "today.tooOld": "This catalog doesn\u2019t know about the daily report yet (404 on /api/today).\nUpdate the web app and try again.",
-  "today.httpError": "The catalog returned {status}. {detail}",
-  "today.nothing": "Nothing new {since}. All yours.",
-  "today.emptyTitle": "Empty catalog: nothing scanned yet.",
-  "today.emptyBody": "Scan a folder of projects to fill it. Scanning runs locally and stores metadata only: your code never leaves this disk.",
-  "today.title": "Today",
-  "today.waiting": "Waiting on you",
-  "today.born": "New in the catalog",
-  /*
-    The first visit has no previous visit, and for weeks it was told the same sentence as everyone
-    else. `visit.json` does not exist until someone has looked, `visitWindow` answers null, and
-    `period` turned that null into "since you last looked" for a person who had never looked — the
-    very first line of the product, and a comparison with nothing. Null is not "a long time ago"
-    either: with no mark, `getDailyReport` opens a window of exactly one day, so this names the
-    window instead of inventing the comparison.
-   */
-  "today.firstLook": "in the last day, on your first look",
-  "today.sinceLastVisit": "since you last looked",
-  "today.sinceToday": "since today at {time}",
-  "today.sinceYesterday": "since yesterday at {time}",
-  "today.sinceWeekday": "since {weekday} at {time}",
-  "today.sinceDate": "since {date}",
-  "today.minutesAgo": "{n} min ago",
-  "today.hoursAgo": "{n} h ago",
-  "today.daysAgo": "{n} day{s} ago",
-  "today.projectsTouched": "{n} project{s} touched",
-  "today.commits": "{n} commit{s}",
-  "today.proposals": "{n} proposal{s}",
-  "today.newProjects": "{n} new project{s}",
-  "today.fromAgents": "{n} from agents",
-  "today.andMoreCommits": "and {n} more commit{s}",
-  "today.verified": "verified",
-  "today.unverified": "unverified",
-  "today.decideAt": "decide them at {url}",
-  "today.notUp": "The catalog is not up:",
-  "today.nobodyAt": "nobody answers at {api}.",
-  "today.startIt": "Start it with",
-  "today.andType": "and type",
-  // ── Does it build? ────────────────────────────────────────────────────────
-  "check.running": "Checking {name}: installs and builds in a separate worktree \u2014 your folder is untouched. This can take a few minutes.",
-  "check.dirty": "You have uncommitted changes: the verdict is about the last commit.",
-  "check.saved": "The verdict is saved on the project's page.",
-  "usage.check": "Usage: panoma check <project>",
-  "open.httpError": "The catalog returned {status}.",
-  "open.noMatch": "No project looks like \u201C{query}\u201D.",
-  "open.noMatchHint": "If it is new, put it in the catalog: panoma scan <path> --save",
-  "open.several": "\u201C{query}\u201D matches {n} projects:",
-  "open.copy": " \xB7 copy",
-  "open.andMore": "and {n} more",
-  "open.useSlug": "Try again with the exact slug.",
-  "open.opened": "{name} opened with {with}",
-  "card.code": "Code",
-  "card.git": "Git",
-  "card.origin": "Origin",
-  "card.icon": "Icon",
-  "card.health": "Health",
-  "card.unsaved": "Unsaved",
-  "card.howToRun": "How to run it",
-  /*
-    A family of two folders holds exactly one copy, which is the commonest first find on any disk:
-    `findDuplicateFamilies` discards a group only when it has fewer than two **members**, and the
-    copies are the members minus the canonical one. So this printed «(1 copies)» — and the plurals
-    guard could not see it, because an exemption there claimed that a group of one copy is never
-    formed. It is formed by every single duplicated pair. The exemption went with this fix.
-   */
-  "card.copies": "({n} cop{ies})",
-  "card.analyzed": "files analyzed: {files} \xB7 in {ms}ms",
-  "card.truncated": " (truncated)",
-  "card.agents": "Agents",
-  "card.run": "Run",
-  "card.needs": "Needs",
-  "card.env": "Env",
-  "card.envMissing": "{file} \xB7 without a value: {n}",
-  "card.risks": "Unsaved",
-  "card.today": "today",
-  "card.yesterday": "yesterday",
-  "card.daysAgo": "{n} day{s} ago",
-  "card.monthsAgo": "{n} month{s} ago",
-  "card.yearsAgo": "{n} year{s} ago",
-  /*
-    The eight risks of unsupported work.
-    They are the SAME words as `apps/web/lib/i18n.ts`, copied on purpose and not similar: the same
-    fact cannot be read differently depending on the screen from which it is viewed. Two codes
-    carry a single form — `unversioned` has no number and `untracked` does not change with it —
-    and `no-commits` changes the entire sentence, not the plural.
-   */
-  "risk.unversioned": "not under version control",
-  "risk.no-commits": "repository with no commits at all",
-  "risk.no-commits.n": "{n} file{s} and not a single commit",
-  "risk.no-remote": "no remote \xB7 {n} commit only on this disk",
-  "risk.no-remote.n": "no remote \xB7 {n} commit{s} only on this disk",
-  "risk.unpushed": "{n} commit not pushed",
-  "risk.unpushed.n": "{n} commit{s} not pushed",
-  "risk.uncommitted": "{n} file not committed",
-  "risk.uncommitted.n": "{n} file{s} not committed",
-  "risk.untracked": "{n} not added to git",
-  "risk.stashes": "{n} stash saved",
-  "risk.stashes.n": "{n} stash{es} saved",
-  "risk.behind": "{n} commit to pull",
-  "risk.behind.n": "{n} commit{s} to pull",
-  "origin.own": "you made it",
-  "origin.forked": "started from someone else\u2019s work",
-  "origin.foreign": "you didn\u2019t start it",
-  "origin.template": "generated from a template",
-  "origin.no-signals": "no way to tell",
-  "families.title": "Copies grouped",
-  "families.main": "main",
-  "families.detected": "Copies found",
-  "families.summary": "{families} famil{fs} \xB7 {folders} duplicate folder{ds} \xB7 {size} of repeated code",
-  "families.alive": "\u2713 alive",
-  "families.copy": "\xB7 copy",
-  "families.noGit": "no git",
-  "families.sameDate": "same date",
-  "families.daysBehind": "{n} d behind",
-  "families.none": "\u2713 No copies of the same project found.",
-  "kind.framework": "Frameworks",
-  "kind.model": "Models",
-  "kind.language": "Languages",
-  "kind.runtime": "Runtimes",
-  "kind.platform": "Platforms",
-  "kind.database": "Data",
-  "kind.tool": "Tools",
-  "kind.package-manager": "Package manager",
-  "severity.critical": "critical",
-  "severity.high": "high",
-  "severity.medium": "medium",
-  "severity.low": "low",
-  "severity.unknown": "unknown",
-  "secret.stripe-live": "Stripe live secret key",
-  "secret.stripe-test": "Stripe test secret key",
-  "secret.aws": "AWS access key",
-  "secret.github-token": "GitHub token",
-  "secret.anthropic": "Anthropic API key",
-  "secret.openai": "OpenAI API key",
-  "secret.slack": "Slack token",
-  "secret.private-key": "Private key",
-  "secret.supabase-service": "Supabase service_role key",
-  "secret.google-api-key": "Google API key",
-  "secret.sendgrid": "SendGrid key",
-  "secret.env-file": "git-tracked .env file",
-  "secret.key-file": "git-tracked key file",
-  "secret.google-service-account": "git-tracked Google service account",
-  "secret.ssh-private-key": "git-tracked SSH private key",
-  "secrets.reading": "Reading the tracked files of every repository\u2026",
-  "secrets.none": "No credential with a recognizable shape in the tracked files.",
-  "scan.none": "No project found under {path}",
-  "scan.noSuchPath": "The folder {path} doesn\u2019t exist. Is the path spelled right?",
-  "scan.notAFolder": "{path} is a file, and scanning works on folders.",
-  "scan.noPermission": "The system wouldn\u2019t let {path} be read. On macOS, Desktop and Documents grant access app by app: System Settings \u2192 Privacy & Security \u2192 Files and Folders, and allow your terminal.",
-  "scan.tasting": "That was a taste: nothing was saved. To save it and see it in one go: npx panoma up {path}",
-  "scan.analyzing": "Analyzing {n} project{s} in {path}\u2026",
-  "scan.failed": "\u2717 {root}: {reason}",
-  "scan.saved": "{n} project{s} in the catalog",
-  "scan.saveFailed": "Couldn\u2019t save to the catalog: {reason}",
-  "scan.wrote": "Wrote {path}",
-  "scan.nextApp": "The next step is the app: npx panoma up",
-  "scan.catalogAt": "Your catalog is at {api}",
-  "scan.projects": "{n} project{s}",
-  /*
-    A pair of keys and not a form gap, because the figure drags three words at once: with a single
-    copy the sentence needs «looks», «a copy» and «it», and no suffix does that to a verb, a noun
-    and a pronoun in the same pass. The singular spells the number out rather than printing it —
-    «1 looks like a copy» would be the same bug wearing the bandage — and it reads after the count
-    that opens the line: «12 projects · one looks like a copy».
-   */
-  "scan.looksLikeCopiesOne": "one looks like a copy \xB7 use -d to see it",
-  "scan.looksLikeCopiesMany": "{n} look like copies \xB7 use -d to see them",
-  "scan.seeFullCard": "use -v for the full report",
-  /*
-    Two keys and not a form hole, because with a commit the noun does not change by itself: the
-    verb changes. '1 commits that live' is the mistake of the number at the end again, and not
-    even `{s}` fixes it — 'that lives' requires the whole sentence.
-   */
-  "scan.noRemote": "{n} with no remote ({commits} commit that lives only on this disk)",
-  "scan.noRemote.n": "{n} with no remote ({commits} commits that live only on this disk)",
-  "scan.unsavedWork": "{n} with unsaved changes",
-  "version.newer": "There is a newer panoma: {latest} (you have {current}).\nUpdate with: npm i -g panoma@latest \u2014 or npx panoma@latest if you use npx.",
-  "enrich.working": "Fetching latest versions and vulnerabilities\u2026",
-  "enrich.done": "{n} package{s} up to date \xB7 {outdated} outdated \xB7 {vulns} with advisories",
-  "enrich.failed": "Couldn\u2019t enrich: {reason}",
-  "disk.regenerate": "regenerate with one command",
-  "disk.missing": "{n} folder{s} in the catalog no longer on disk",
-  "search.matches": "{n} match{es}",
-  "search.inProjects": "in {n} project{s}",
-  "secrets.checked": "{n} repositor{y} checked \xB7 {skipped} without git \xB7 ",
-  "secrets.publicByDesign": "{n} skipped as public by design",
-  "secrets.revokeFirst": "Revoke it in the provider\u2019s dashboard first: that is the only thing that turns it off.\nClean the history afterwards, with git filter-repo. Deleting the file and committing\ndoes nothing \u2014 the key is still in every earlier commit.",
-  "secrets.findings": "{n} finding{s}",
-  "secrets.inProjects": "in {n} project{s}",
-  "agentKey.key": "Key",
-  "agentKey.onlyNow": "(shown only now; the catalog stores only its hash)",
-  "mcp.written": "MCP configuration written to {path}",
-  "mcp.merged": "MCP configuration merged into {path}",
-  "mcp.coexists": "coexists with: {list}",
-  "mcp.gitWarning": "That file holds the key in the clear, and git would carry it.\n      Before you commit: echo '{name}' >> .gitignore",
-  "disk.title": "Disk",
-  "disk.inProjects": "{size} across {n} project{s}",
-  "disk.breakdown": "The per-project breakdown, at {url}",
-  "enrich.registries": "Registries",
-  "enrich.resolved": "{n} package{s} resolved out of {checked} queried",
-  "enrich.unresolvable": "{n} unpublished (private, local or renamed)",
-  "enrich.retry": "{n} failed and will be retried",
-  "enrich.portfolio": "Portfolio state",
-  /*
-    «dependency» and «advisory» lose the y when they grow, so the gap carries both halves, the same
-    shape `cli.catalogUpdated` uses for «technology» and «family». One stale package used to read
-    «1 direct dependencies outdated», and one stale package is exactly what a portfolio of one
-    project has.
-   */
-  "enrich.outdated": "{n} direct dependenc{ies} outdated",
-  "enrich.advisories": "{n} security advisor{ies}",
-  /*
-    A participle and not a conjugated verb. This half is printed right after `enrich.advisories`
-    and its subject is that other count, which this sentence never receives: with one advisory it
-    said «1 security advisory affect 2 versions in use». «affecting» has no number to agree with,
-    so the seam between the two keys stops being a place where the grammar can break.
-   */
-  "enrich.affect": "affecting {n} version{s} in use",
-  "enrich.noVulns": "no known vulnerabilities",
-  "disk.measuring": "Measuring each project\u2019s disk usage\u2026",
-  "disk.total": "{total} in total \xB7 {reclaimable} regenerates itself",
-  "search.none": "Nothing matches \u201C{query}\u201D.",
-  "search.found": "{n} match{s} in {projects} project{ps}",
-  "search.searched": "{n} repositor{y} searched",
-  "search.skippedNoGit": "{n} with no git, couldn\u2019t be searched",
-  "describe.asking": "Asking the model\u2026",
-  "run.queued": "Proposal queued \xB7 {id}",
-  "run.watch": "Follow it with panoma, or on the project\u2019s card.",
-  "run.isolating": "Isolating {slug} in a worktree, installing and running the tests\u2026",
-  "run.alreadyFailed": "Already tried, already failed",
-  "run.runId": "run {id}",
-  "run.hardened": "clean environment, no container",
-  "run.local": "not isolated",
-  "agentKey.created": "Key created for {name}",
-  "agentKey.installed": "Written to {path}",
-  "agentKey.keep": "Keep it: it is not shown again.",
-  "agentKey.registered": "Agent {name} registered.",
-  "ai.unknownSubHint": "Use: panoma ai [status] \xB7 use <provider> \xB7 key <provider> \xB7 ask <question>",
-  "ai.defaultModel": "default model",
-  "ai.useSubscription": "Use a subscription you already have",
-  "ai.useSubscriptionNote": "(no keys; Panoma calls the agent)",
-  "ai.installed": "installed",
-  "ai.agentBroken": "installed, but it does not run",
-  "ai.notInstalled": "not installed",
-  "ai.onDemand": "Panoma never calls a model on its own: only when you ask it to.",
-  /*
-    Why a model cannot be connected, and what to do — in English, like the whole terminal.
-    `resolveCredential` launches these two errors in Spanish, and here they were printed as is:
-    the only line that says how to get out of the problem came out in another language. It is not
-    a word from a provider —there is not one to quote yet—, Panoma says it. The web solved it the
-    same way and earlier, in `lib/model-errors.ts`: `NoCredentialError` is recognized, which
-    travels typed with the entire provider inside, and the remedy is drafted here. What is indeed
-    someone else's word —a 429, the network down— is left as is, because translating it would be
-    inventing.
-   */
-  "ai.noCredential": "the {name} credential is missing",
-  "ai.hintCli": "Install {name} and sign in; Panoma will call \u201C{command}\u201D.",
-  "ai.hintOauth": "Sign in to {name} with: panoma ai use {id}",
-  "ai.hintKey": "Save the key with: panoma ai key {id}   Get one at {url}",
-  "ai.missingProvider": "The provider is missing.",
-  "ai.available": "Available: {list}",
-  "ai.chosenProvider": "Provider: {name}",
-  "ai.chosenModel": " \xB7 model {model}",
-  "ai.sessionModel": "from the session",
-  "ai.keyUsage": "Usage: panoma ai key <provider>",
-  "ai.keyProviders": "With a key: {list}",
-  "ai.pasteKey": "Paste the {name} key and press Enter",
-  "ai.pasteKeyNote": "(it won\u2019t echo)",
-  "ai.saved": "Saved {masked} for {name} at {path}",
-  "ai.keyDetail": "key {masked} ({source})",
-  "ai.fromEnv": "{name} in the environment",
-  "ai.storedKey": "stored {masked}",
-  "ai.source.env": "environment",
-  "ai.source.file": "file",
-  "ai.source.agent-session": "agent session",
-  "ai.source.login": "logged in",
-  "ai.activeProvider": "Active provider",
-  "ai.none": "none \u2014 pick one with 'panoma ai use <provider>'",
-  "ai.withKey": "Connect with an API key",
-  "ai.withSession": "With a session you already have",
-  "ai.configAt": "Config at {path} (mode 0600).",
-  "ai.unknownProvider": "Unknown provider: {id}",
-  "ai.unknownSub": "Unknown subcommand: {sub}",
-  "ai.chosen": "Panoma now thinks with {name}.",
-  "ai.keyPrompt": "Paste the {name} key and press \u21B5 (it won\u2019t echo):",
-  "ai.keyEmpty": "No key was read.",
-  "ai.keySaved": "{name} key saved to {path}.",
-  "ai.askUsage": 'Usage: panoma ai ask "your question"',
-  "ai.answeredBy": "\u2014 {model}",
-  "hooks.noRepo": "There is no git repository here ({root}).",
-  "hooks.noRepoHint": "Hooks are installed inside the project you want to record.",
-  "hooks.foreignPostCommit": "There is already a post-commit that isn\u2019t Panoma\u2019s:",
-  "hooks.foreignHint": "Leaving it alone. If you want both, add this line at the end by hand:",
-  "hooks.stopInstalled": "Claude Code Stop hook {path}",
-  "hooks.signalInstalled": "Claude Code PreToolUse signal \u2014 sleeping notes fire on their paths {path}",
-  "hooks.updated": " (updated)",
-  "hooks.removeWith": "remove them with: panoma hooks --remove",
-  "hooks.badJson": "{path} is not valid JSON ({reason}). Leaving it as it is.",
-  "hooks.cantWrite": "Couldn\u2019t touch {path}: {reason}",
-  "hooks.noPanomaHook": "\xB7 {path} had no Panoma hook",
-  "hooks.statusTitle": "Passive capture in",
-  "hooks.gitPostCommit": "git post-commit {path}",
-  "hooks.noSettings": "\xB7 no .claude/settings.json in this project",
-  "hooks.statusHint": "panoma hooks --install to add them \xB7 --remove to take them out",
-  "hooks.installed": "Stop hook installed {path}",
-  "hooks.postCommit": "post-commit installed {path}",
-  "hooks.notOurs": "the post-commit is not Panoma\u2019s: leaving it alone",
-  "hooks.postCommitRemoved": "post-commit removed {path}",
-  "hooks.stopRemoved": "Stop hook removed {path}",
-  "hooks.cantTouch": "Couldn\u2019t touch {path}: {reason}",
-  "hooks.notJson": "not a JSON object",
-  "hooks.hooksNotObject": "\u201Chooks\u201D is not an object",
-  "hooks.stopNotList": "\u201Chooks.Stop\u201D is not a list",
-  "mcp.badJson": "{path} is not valid JSON ({reason}). Leaving it alone.",
-  "env.notBuilt": "panoma is not on the PATH and the CLI is not built: this points at {entry}, which only very recent Node versions can run. Build it with \u201Cpnpm --filter panoma build\u201D and install again.",
-  "mcp.installed": "Panoma registered in {path}",
-  "mcp.replaced": "The previous panoma entry was replaced.",
-  "mcp.notBuilt": "The MCP server isn\u2019t built yet ({path}).\nBuild it with: pnpm --filter @panoma/mcp build",
-  "mcp.notAnObject": "The .mcp.json that\u2019s there isn\u2019t a JSON object. Leaving it alone.",
-  "mcp.badToml": "{path} has a syntax error. Leaving it alone: fix it and paste this yourself.",
-  "mcp.tomlManual": "Panoma is already in {path}, written your way. Leaving it alone: update it yourself with this.",
-  "mcp.serversNotAnObject": "The .mcp.json that\u2019s there has an \u201CmcpServers\u201D that isn\u2019t an object. Leaving it alone.",
-  "error.unknownCommand": "Unknown command: {command}",
-  "error.noCommand": "(none)",
-  "error.unknownFlag": "Unknown option: {flag}",
-  "error.didYouMean": "did you mean {guess}?",
-  "error.unknownIsolation": "Unknown isolation level: {value}",
-  "error.seeHelp": "panoma --help to see them all.",
-  "error.needsValue": "{flag} needs a value.",
-  "error.isolationLevels": "The ones there are: {list}",
-  "error.badLimit": "--limit needs a whole number greater than zero.",
-  "error.folderAndTerminal": "--folder and --terminal ask for two different things. Pick one.",
-  "error.installAndRemove": "--install and --remove ask for two different things. Pick one.",
-  // ── The agents' .md ───────────────────────────────────────────────────────
-  "md.usage": "panoma md [check|fix|init|sync|review] [path]",
-  "md.unknownSub": "Unknown subcommand: {sub}",
-  "md.unknownSubHint": "Available: check (audit), fix (repair the obvious), init (add the block), sync (regenerate it), review (model opinion, paid).",
-  "md.checking": "Checking the instructions file against the disk\u2026",
-  "md.noDocs": "There is no AGENTS.md or CLAUDE.md here.",
-  "md.noDocsHint": "panoma md init creates AGENTS.md with panoma's context block.",
-  "md.fileHead": "{file} \xB7 {tokens} token{ts} \xB7 {lines} line{ls}",
-  "md.managedTag": "has panoma's block",
-  "md.clean": "everything it claims exists",
-  "md.findingCount": "{n} claim{s} no longer true",
-  "md.findingCountOne": "1 claim that is no longer true",
-  "md.blockBroken": "the panoma block never closes: fix it by hand",
-  "md.versionWrong": "the project runs {v}",
-  "md.envMissing": "the env example does not declare it",
-  "md.envNear": "not declared in the env example; there is {names}",
-  "md.fixDone": "{n} fix{es} in {file} \xB7 {m} left for your hand",
-  "md.fixClean": "{file}: nothing fixable automatically; what remains needs your hand.",
-  "md.fixNothing": "Nothing to fix: everything it claims exists.",
-  "md.reviewAsking": "Reading the instructions file and asking the model\u2026",
-  "md.reviewBy": "opinion by {model} on \u201C{name}\u201D \xB7 stored in its page",
-  "md.inherited": "also inherits: {path} ({tokens} tokens)",
-  "md.inheritedHint": "agents read it on top of the project's own; check it with: panoma md check {dir}",
-  "md.findingRow": "line {line} \xB7 {claim} \u2014 {reason}",
-  "md.pathMissing": "does not exist in the project",
-  "md.pathMovedTo": "missing; there is one at {path}",
-  "md.scriptMissing": "not in the package.json scripts",
-  "md.scriptNear": "not in the scripts; there is {names}",
-  "md.truncated": "the project has more files than the index holds: paths were not checked",
-  "md.touchedBy": "last agent touch: {agent} on {file} (+{added} \u2212{deleted})",
-  "md.notCataloged": "The project is not in the catalog: the block only carries what the disk shows.",
-  "md.notCatalogedHint": "panoma scan . --save registers it; the block gains dependencies, advisories and tasks.",
-  "md.initDone": "Panoma block written to {file}.",
-  "md.initCreated": "There was no instructions file: {file} created with the block.",
-  "md.initKeeps": "It regenerates itself while the watcher runs, or by hand with: panoma md sync",
-  "md.bridgeCreated": "CLAUDE.md created with the @AGENTS.md import: Claude Code reads only its own file.",
-  "md.bridgeMissing": "Claude Code will not load this file: it reads CLAUDE.md, never AGENTS.md.",
-  "md.bridgeMissingHint": "panoma md init writes a one-line CLAUDE.md that imports AGENTS.md.",
-  "md.syncDone": "Block regenerated in {file}.",
-  "md.syncSame": "The block in {file} was already up to date: nothing changed.",
-  "md.syncNone": "There is no panoma block in this project.",
-  "md.syncNoneHint": "Add it with: panoma md init",
-  "md.broken": "The block is broken: {reason}",
-  "cli.httpError": "The catalog returned {status}. {detail}",
-  "cli.ingestRejected": "The catalog rejected the ingest ({status}). {detail}",
-  /*
-    A pair each, because the word the number drags here is a verb: «1 were skipped» is the
-    number-at-the-end bug in its worst shape, and a noun suffix never reaches «was». The noun that
-    was missing goes in too — a bare figure followed by a verb left the reader guessing what had
-    been retired.
-   */
-  "cli.removedOne": "{n} project no longer existed and was retired",
-  "cli.removedMany": "{n} projects no longer existed and were retired",
-  /*
-    The three sentences npx needs, and none of them says «trial».
-    What runs under `npx panoma` is this same package, byte for byte: nothing is limited and
-    nothing expires. What is temporary is the command, and saying anything stronger would be a
-    lie the rest of this product does not tell.
-   */
-  "npx.hooksRefused": "Hooks need a panoma that is still here tomorrow.",
-  "npx.hooksRefusedWhy": "This copy runs from npx, which keeps it for one command and then lets it go. A hook written\n  now would call a command that stops existing the moment this one ends \u2014 and hooks stay quiet\n  by design, so nothing would ever tell you.",
-  "npx.hooksRefusedHow": "Install it and try again:  npm i -g panoma",
-  /*
-    And the same refusal for the MCP, which needed it for the same reason and did not have it.
-    `--install` writes a permanent file in somebody's agent that names a path inside npx's cache.
-    That path is alive today and gone whenever npm clears it, and an MCP server that fails to
-    start is not an error anyone sees: the agent simply comes up without the tools.
-   */
-  "npx.mcpRefused": "An agent's configuration outlives the copy that would write it.",
-  "npx.mcpRefusedWhy": "This copy runs from npx, which keeps it for one command and then lets it go. The configuration\n  written now would point inside that cache, so the agent would stop getting the tools the day it\n  is cleared \u2014 and it would come up without them saying nothing, which is how MCP fails.",
-  "npx.mcpRefusedHow": "Install it and try again:  npm i -g panoma",
-  "npx.upEphemeral": "running from npx \xB7 to keep the command: npm i -g panoma",
-  "cli.catalogUpdated": "Catalog updated: {projects} project{ps}, {technologies} technolog{ts}, {packages} package{ks}, {families} famil{fs}",
-  /*
-    «as you asked» is literal and was checked before it was kept: the rows come from the
-    `exclusions` table, `excludeProject` writes them when someone takes a project out of the
-    catalog from the web —typing its name to confirm— and `ingestPortfolio` reads them to skip
-    those roots. The CLI never writes there; it only reports what the ingest left out.
-   */
-  "cli.excludedOne": "{n} project was skipped for living outside the catalog, as you asked",
-  "cli.excludedMany": "{n} projects were skipped for living outside the catalog, as you asked",
-  "cli.reslugged": "{n} changed address when unique names were handed out",
-  "search.nothingTracked": "No git-tracked file contains \u201C{term}\u201D.",
-  "disk.walking": "Walking each project\u2019s tree. This takes several minutes\u2026",
-  "secrets.revoke": "Revoke it in the provider\u2019s dashboard first: that is the only thing that disables it.\nClean the history afterwards, with git filter-repo. Deleting the file and committing\ndoes not help \u2014 the key is still in every earlier commit.",
-  "describe.reading": "Reading {slug} and asking the model\u2026",
-  "enrich.asking": "Querying public registries and OSV.dev\u2026",
-  "mcp.configTitle": "MCP configuration",
-  "mcp.pasteIt": "Paste it into .mcp.json or ~/.claude.json \u2014 or run it again with --install",
-  "mcp.cannotWrite": "Nothing written: this agent keeps its MCP servers in a format I will not touch. Paste it into {path} yourself",
-  "mcp.unknownAgent": "Nothing written: I don\u2019t know where this agent keeps its MCP servers. Paste it wherever it does.",
-  "mcp.restart": "Restart {name} so it picks this up.",
-  "mcp.updated": "the \u201Cpanoma\u201D entry that was already there has been updated",
-  "run.securityFix": "Security fix",
-  "run.isolation": "Isolation",
-  "run.container": "ephemeral container",
-  "run.steps": "Steps",
-  "run.notApplied": "The change is NOT applied in your folder.",
-  "run.reviewAt": "Review the patch and decide at {url}",
-  "run.nothingToChange": "Nothing to change.",
-  "run.failed": "Failed",
-  "run.proposalVerified": "Proposal verified",
-  "run.proposalUnverified": "Proposal NOT verified",
-  "run.branch": "branch",
-  "open.openedWith": "opened with {tool}",
-  "describe.writtenBy": "written by {model}",
-  "run.stepOutput": "output of \u201C{step}\u201D:",
-  "cli.trace": "PANOMA_DEBUG=1 to see the full trace.",
-  "card.distribution": "Distribution",
-  "card.noLockfile": "no lockfile",
-  "card.lockUnresolved": "{path} (versions unresolved)",
-  "hooks.noPostCommit": "\xB7 there was no post-commit",
-  "ai.notEncrypted": "The file is mode 0600, but it is not encrypted.",
-  "mcp.noMonorepo": "Can\u2019t find the monorepo from here, so the config points at @panoma/mcp,\nwhich is not published on npm yet. Run this from your clone of the repository.",
-  "twin.usage": "panoma twin [sources|allow|revoke|forget|mine|verdicts|distill|synthesize|taste|score|design|look]",
-  "twin.unknownSub": "Unknown subcommand: {sub}",
-  "twin.unknownSubHint": "The ones that exist: sources (which histories are here and which are permitted), allow and revoke (grant and take back that permission), forget (delete what was stored), mine (what reading them yields), verdicts (what ended up in the catalog), distill (read them and pull out observations), synthesize (turn those observations into your portrait), taste (the portrait), score (how often you correct it) and look (what\u2019s wrong with a screen you were just handed).",
-  "twin.needsSource": "Missing the history: panoma twin {sub} <source>",
-  "twin.badSource": "\u201C{source}\u201D is not an agent history Panoma knows about.",
-  "twin.badSourceList": "The ones that exist: {list}",
-  "twin.sourcesTitle": "Agent histories on this machine",
-  "twin.sourcePresent": "{files} file{s} \xB7 {size}",
-  "twin.sourceAbsent": "not on this machine",
-  "twin.consentAllowed": "permitted",
-  "twin.consentAllowedHint": "take it back with panoma twin revoke {source}",
-  "twin.consentDenied": "not permitted",
-  "twin.consentDeniedHint": "grant it with panoma twin allow {source}",
-  "twin.consentNoReader": "no reader yet",
-  "twin.consentNoReaderHint": "this format cannot be read yet: the permission would open nothing",
-  "twin.sourcesTotal": "{n} source{s} on disk \xB7 {files} file{fs} \xB7 {size}",
-  "twin.sourcesAllowed": "{n} permitted \xB7 {files} file{fs} \xB7 {size} that panoma twin mine would actually open",
-  "twin.sourcesNoneAllowed": "None of them is permitted yet, so panoma twin mine would open nothing.",
-  "twin.sourcesNone": "There are no agent histories on this machine.",
-  "twin.sourcesNoneHint": "Twin reads what Claude Code, Codex, Cursor or Aider already wrote while you\nworked. With none of them here there is nothing to read.",
-  "twin.sourcesNext": "panoma twin allow <source> grants the permission; panoma twin mine reads the permitted ones.",
-  "twin.nothingRead": "Only the size was measured here: not one file opened, not one byte stored.",
-  "twin.granted": "Permission granted: {label}",
-  "twin.grantedCovers": "covers {files} file{s} \xB7 {size} in {path}",
-  "twin.grantedAbsent": "nothing there for now: that tool has not written on this machine",
-  "twin.grantedNoReader": "This format cannot be read yet, so the permission is stored with nothing to open.",
-  "twin.grantedNext": "panoma twin mine --source {source} reads it now.",
-  "twin.revoked": "Permission taken back: {label}",
-  "twin.revokedDetail": "Those files do not get opened again. What was printed back then was not stored\n  anywhere either, other than whatever you sent to the catalog with --save.",
-  "twin.mineReading": "Reading {label} on this disk\u2026",
-  "twin.mineTitle": "What comes out of your histories with the agents",
-  "twin.mineNothingAllowed": "No history is permitted yet, so not one file has been opened.",
-  "twin.mineNothingAllowedHint": "panoma twin sources shows which ones are here and how big they are, and that is where\n  the identifier panoma twin allow <source> wants comes from.",
-  "twin.mineDenied": "{label} is on disk and is not permitted, so it was not opened.",
-  "twin.mineDeniedHint": "panoma twin allow {source} grants it.",
-  "twin.mineNoReader": "{label} gets measured, but cannot be read yet.",
-  "twin.mineNoReaderHint": "Its format has no reader yet, and promising one on this screen would be a lie.",
-  "twin.projectFilter": "Only sessions whose directory starts with \u201C{prefix}\u201D. For now --project filters by\n  path, and not by the catalog entry.",
-  "twin.funnelTitle": "The funnel, all of it",
-  "twin.funnelRead": "{files} file{s} read \xB7 {size}",
-  "twin.funnelSessions": "sessions",
-  "twin.funnelUserTurns": "turns of yours, with none of the above still in",
-  "twin.funnelToolResults": "tool results, which you did not write",
-  "twin.funnelSidechain": "subagent turns, which you did not either",
-  "twin.funnelCommands": "slash commands: instructions, not reactions",
-  "twin.funnelReactions": "reactions to something you had been handed",
-  "twin.funnelBriefs": "one line or shorter",
-  "twin.funnelSpontaneous": "with nobody asking you anything",
-  "twin.funnelWithSignal": "carrying a signal we can name",
-  "twin.funnelWholeCorpus": "The funnel counts the whole history: --project filters the reactions shown,\n    not the files that get read.",
-  "twin.mineTotal": "{n} histor{y} together \xB7 {files} file{fs} \xB7 {size} \xB7 reactions: {reactions} \xB7 with a signal: {signals}",
-  "twin.samplesTitle": "A few of them, exactly as you wrote them",
-  "twin.sampleQuote": "\u201C{text}\u201D",
-  "twin.sampleBrief": "one line",
-  "twin.sampleRedacted": "something covered up, in case it was a credential",
-  "twin.samplesNone": "No reaction of that shape in this history.",
-  "twin.samplesMore": "and {n} more that do not fit here \xB7 --limit decides how many are shown",
-  "twin.saveSending": "Sending what came out to the catalog, in batches of {size}\u2026",
-  "twin.saved": "stored: {saved} \xB7 already there: {duplicates} \xB7 no project: {unmatched}",
-  "twin.savedProjects": "spread across {n} catalogued project{s}",
-  "twin.savedUndated": "and {n} with no timestamp in the transcript, not stored and not fixable",
-  "twin.savedUnmatched": "\u201Cno project\u201D means its folder is not catalogued: panoma scan <path> adds it, and they\n    attach on the next save.",
-  "twin.saveNothing": "No reaction came out, so there is nothing to send.",
-  "twin.saveRejected": "The catalog rejected the reactions ({status}). {detail}",
-  "twin.forgetUsage": "Say what to forget: panoma twin forget <source>. The ones that work:",
-  "twin.forgotten": "Forgotten: {n} verdict{s} deleted from the catalog.",
-  "twin.nothingSaved": "None of this was stored: it is read from disk, printed, and that is the end of it.",
-  "twin.saveHint": "--save stores them in the catalog, the only thing that takes them out of here.",
-  "twin.verdictsTitle": "What is already stored in the catalog",
-  "twin.verdictsShown": "verdicts: {shown} of {total}, grouped by project",
-  "twin.verdictsFiledBy": "The project is the identity each one was filed under \u2014its repository\u2019s root commit,\n  plus its path inside it\u2014, and not the folder\u2019s name: that way it survives a rename.",
-  "twin.verdictsIn": "{n} verdict{s}",
-  "twin.verdictsMore": "and {n} more in the catalog \xB7 --limit decides how many are fetched",
-  "twin.verdictsSource": "Only the ones from \u201C{source}\u201D.",
-  "twin.verdictsNone": "The catalog has no stored verdicts.",
-  "twin.verdictsNoneSource": "No stored verdict came out of \u201C{source}\u201D.",
-  "twin.verdictsNoneHint": "panoma twin mine --save stores them, the only thing that takes them off the disk.",
-  "twin.verdictsRejected": "The catalog could not list the verdicts ({status}). {detail}",
-  "twin.distillTitle": "Distil your verdicts into statements about your taste",
-  "twin.distillEstimate": "{verdicts} verdict{s} \xB7 about {tokens} tokens \xB7 {provider} \xB7 {model}",
-  "twin.distillCost": "That is what it would cost, counted before spending it.",
-  "twin.distillDryRun": "With --dry-run it stops here: nothing was sent to any model.",
-  "twin.distillDryRunHint": "Without --dry-run it runs, and whatever comes back waits for your review.",
-  "twin.distillRunning": "Reading those verdicts with the model\u2026",
-  "twin.distillNothing": "There are no verdicts to distil.",
-  "twin.distillNothingHint": "panoma twin mine --save fills the catalog, and panoma twin verdicts shows what is in it.",
-  "twin.distilled": "observations: {observed} \xB7 stored: {saved} \xB7 out of {verdicts} verdict{s}",
-  "twin.distilledUsage": "{model} \xB7 {input}\u2192{output} tokens",
-  "twin.distilledModel": "written by {model}",
-  "twin.distilledNone": "The model pulled no observation out of these quotes.",
-  "twin.distilledNext": "panoma twin synthesize turns them into your portrait, without asking you anything.",
-  "twin.distillMore": "still unread: {left} \xB7 next pass, number {pass}\u2026",
-  "twin.distillPasses": "chained passes: {n}",
-  "twin.distillNoModel": "No model is connected, so there is nothing to distil with.",
-  "twin.distillNoModelHint": "panoma ai shows which ones you can connect \xB7 panoma ai use <provider> picks one.",
-  "twin.distillRejected": "The catalog could not distil ({status}). {detail}",
-  "twin.tasteFull": "It doesn\u2019t fit: the portrait would take {chars} of {cap} characters, and nothing was saved.",
-  "twin.tasteFullHint": "For a new belief to get in, another has to come out \u2014 or stay only in its project,\n  which is what scoping it in the catalog does.",
-  "twin.synthTitle": "Writing your portrait",
-  "twin.synthSorted": "sorted by topic: {n}",
-  "twin.synthMinted": "and {n} new topic that wasn\u2019t on the list",
-  "twin.synthEstimate": "topics: {topics} \xB7 observations: {observations} \xB7 about {tokens} input tokens",
-  "twin.synthModel": "with {provider} \xB7 {model}",
-  "twin.synthRunning": "Synthesizing\u2026",
-  "twin.synthSame": "Nothing changed: the evidence says what it said last time.",
-  "twin.synthDone": "new: {created} \xB7 refined: {refined} \xB7 retired: {retired}",
-  "twin.synthProposed": "{n} belief{s} you signed have a proposed version waiting for your answer.",
-  "twin.synthNext": "panoma twin taste shows how it turned out.",
-  "twin.churnTitle": "How your portrait has moved",
-  "twin.churnMonth": "{month} \u2014 new: {created} \xB7 refined: {refined} \xB7 retired: {retired}",
-  "twin.churnStill": "It hasn\u2019t moved this month: what is there is already said.",
-  "twin.churnOnlyRefined": "This month only rewrote what was already there: nothing new and nothing retired.",
-  "twin.synthNothing": "There is no evidence to synthesize yet.",
-  "twin.synthNothingHint": "panoma twin distill reads your verdicts and pulls out the observations the portrait\n  comes from.",
-  "twin.synthUpToDate": "The portrait is already up to date: no new evidence has come in.",
-  "twin.synthUpToDateHint": "panoma twin distill reads whatever is left of your history, and panoma twin taste shows\n  how the portrait turned out.",
-  "twin.tasteTitle": "Your portrait, exactly as it is written",
-  "twin.tasteBudget": "{chars} of {cap} characters \xB7 {left} left",
-  "twin.tasteCitations": "{n} quote{s}",
-  "twin.tasteFile": "It is written to TASTE.md, in Panoma\u2019s home, so an agent can read it.",
-  "twin.tasteForming": "and {n} belief{s} still forming \xB7 they need more evidence to reach the file",
-  "twin.tasteWaiting": "There are beliefs with evidence to spare that aren\u2019t here because you haven\u2019t let them down: {n}",
-  "twin.tasteWaitingHint": "What the machine works out on its own doesn\u2019t reach your agents until you say yes once,\n  in the catalog. Until then, this is exactly what you signed.",
-  "twin.tasteOnly": "only in {project}",
-  "twin.tasteEmpty": "The portrait is empty: no belief has made it in yet.",
-  "twin.tasteEmptyHint": "panoma twin distill reads your verdicts and panoma twin synthesize writes the portrait.",
-  "twin.tasteRejected": "The catalog could not read the portrait ({status}). {detail}",
-  "twin.topicDesign": "on design",
-  "twin.topicFrontend": "on the interface itself",
-  "twin.topicBackend": "on the server and its data",
-  "twin.topicCli": "on the terminal",
-  "twin.topicTesting": "on how things are checked",
-  "twin.topicCopy": "on the words",
-  "twin.topicWorkflow": "on how you work with your agents",
-  "twin.topicTooling": "on the tooling",
-  "twin.topicData": "on the data",
-  "twin.topicOther": "on everything else",
-  "twin.scoreTitle": "How often you correct it",
-  "twin.scoreCounts": "beliefs: {beliefs} \xB7 forming: {forming} \xB7 signed: {signed}",
-  "twin.scoreCorrections": "you have corrected {corrections} of {shown}",
-  "twin.scoreRate": "{rate}% of what it told you, you had to correct",
-  "twin.scoreDensity": "observations per belief: {density} \xB7 in total: {observations}",
-  "today.critic": "The critic looked on its own",
-  "today.criticLooks": "things it saw in inbox screenshots: {n} \xB7 screenshots looked at: {shots}",
-  "today.criticReviews": "things it saw reading folders: {n} \xB7 folders reviewed: {projects}",
-  "today.criticWhere": "the full verdict is in the catalog, on the critic's screen",
-  "twin.scoreBriefs": "of what the critic has seen you have assigned {ordered} of {findings}",
-  "twin.scoreBriefsRate": "{rate}% of what it points at works for you",
-  "twin.scoreBriefsLaunched": "of those, sent to an agent: {launched}",
-  "twin.scoreBriefsDiscarded": "and you said no to: {discarded}",
-  "twin.scoreReach": "Your portrait goes down to the .md of these projects: {reached} of {projects}",
-  "twin.scoreReachHow": "In none of them is the channel open, so there your agents work knowing nothing about this. You open it inside the folder, with: panoma md init",
-  "twin.designTitle": "What yours looks like",
-  "twin.designFrom": "From projects the critic has read, copies aside: {read} \xB7 of those, with something to look at: {withUi}",
-  "twin.designEmpty": "No fingerprint has been stored yet. The mechanical critic writes one whenever it reviews a folder, and it reviews when there is a commit newer than its last review.",
-  "twin.designProjects": "projects: {projects}",
-  "twin.designFonts": "Typefaces: {fonts}",
-  "twin.designRadii": "Corners: {radii}",
-  "twin.designTraits": "With dark mode: {dark} \xB7 with animation: {animation}",
-  "twin.designRejected": "The catalog would not give the visual portrait: {detail}",
-  "twin.scoreBriefsRelaunched": "some more than once \u2014 launches: {launches}",
-  "twin.scoreGraveyard": "What you veto is not deleted: it stays as negative evidence so it is never proposed\n  again.",
-  "twin.scoreTooFew": "It has told you {shown}, and it takes {floor} for a percentage to say anything. Below\n  that, a single correction moves it more than five points, so it would be describing the\n  last belief you looked at and not your taste.",
-  "twin.scoreNoTrend": "Where it stands today can be said; whether it improves, not yet: neither of the two\n  settled months reaches {floor} beliefs. The current month doesn\u2019t count: its beliefs\n  haven\u2019t been looked at yet.",
-  "twin.scoreBetter": "Of what it told you last month you corrected {recent}%, and of the month before\n  {previous}%: it goes down, which is the only way this has of saying it is learning.",
-  "twin.scoreNotBetter": "Of what it told you last month you corrected {recent}%, and of the month before\n  {previous}%: it does not go down. Until it goes down month over month, the twin is not\n  learning, and this scoreboard will not pretend otherwise.",
-  "twin.scoreNothing": "There is nothing to score yet: no belief has been written.",
-  "twin.scoreNothingHint": "panoma twin distill reads your verdicts and panoma twin synthesize writes the portrait;\n  this number comes out of what you correct there.",
-  "twin.scoreRejected": "The catalog could not count your decisions ({status}). {detail}",
-  "twin.saveRemapped": "{n} quote{s} that were already stored moved to another project: attribution is computed, and this pass resolved it better.",
-  "twin.saveRestated": "Observations that moved project along with their quotes: {n}. Look at them with panoma twin taste.",
-  "twin.corpusLeft": "You have {read} of {total} verdict{s} distilled from your history: {left} still unread,\n  and every pass moves on to the ones that haven\u2019t been used yet.",
-  /*
-    A pair and not a gap: with one verdict the article, the noun and the verb all move at once
-    —«all 1 stored verdicts already back» — and a suffix reaches none of them. The web half of this
-    same sentence was split the same way on the same day.
-   */
-  "twin.corpusDoneOne": "Your whole history has been distilled: the one stored verdict already backs a\n  statement.",
-  "twin.corpusDoneMany": "Your whole history has been distilled: all {total} stored verdicts already back some\n  statement.",
-  "twin.lookNeedsArgs": "The project is missing: panoma twin look <project> [screenshot.png]",
-  "twin.lookNoInbox": "This project has no inbox: {dir} isn\u2019t there.",
-  "twin.lookNoInboxHint": "panoma md init creates it and tells your agents to leave screenshots of what they\n  build in it. The folder is git-ignored.",
-  "twin.lookInboxEmpty": "The inbox is set up and empty: {dir}.",
-  "twin.lookInboxEmptyHint": "Your agents already have the instruction in AGENTS.md. Ask the one that is working to\n  leave a screenshot of the screen it just touched.",
-  "twin.lookFromInbox": "from the inbox, left {when}",
-  "twin.lookInboxMore": "{n} more unseen",
-  "twin.lookInboxSkipped": "files that aren\u2019t images: {n}",
-  "twin.lookTitle": "What\u2019s wrong with this screen",
-  "twin.lookSending": "{file} \xB7 {size}{dims}",
-  "twin.lookNotRedacted": "An image can\u2019t be redacted: it leaves your disk exactly as it is, with whatever is\n  written on it. Look at it before you send it.",
-  "twin.lookTiny": "It is {width} px wide. Below {floor}, whatever gets judged will be about the scale the\n  file was saved at rather than about your screen.",
-  "twin.lookEstimate": "{statements} statement{s} from your portrait \xB7 {tokens} tokens of text \xB7 {size} of image \xB7 {provider}/{model}",
-  "twin.lookCost": "The image is billed separately and every provider counts it its own way: this is its\n  size, not its price.",
-  "twin.lookDryRun": "That\u2019s as far as the rehearsal goes: nothing was looked at.",
-  "twin.lookDryRunHint": "Drop --dry-run to look at it for real.",
-  "twin.lookRunning": "Looking at the screenshot\u2026",
-  "twin.lookClean": "Nothing you have approved is broken on this screen.",
-  "twin.lookBreaks": "Breaks your portrait in {n} place{s}",
-  "twin.lookFix": "ask for: {fix}",
-  "twin.lookAgainst": "breaks: \u201C{statement}\u201D",
-  "twin.lookDropped": "{n} judgement{s} with nothing behind them, dropped.",
-  "twin.lookDroppedHint": "Only what breaks a belief already written comes out here. The more your portrait has,\n  the more it can see: panoma twin synthesize.",
-  "twin.lookUnreadable": "The model didn\u2019t answer in the shape it was asked for, so there is nothing to show. The\n  call still counts against today\u2019s looks.",
-  "twin.lookFooter": "{statements} statement{s} in your portrait \xB7 looks today: {used} of {cap}",
-  "twin.lookSpend": "Today, on looks: {input} input tokens and {output} output.{unmetered}",
-  "twin.lookUnmetered": " {n} unmeasured: that provider doesn\u2019t publish usage.",
-  "twin.lookMissing": "I can\u2019t find that file: {path}",
-  "twin.lookEmpty": "That file is empty: {path}",
-  "twin.lookNotImage": "That isn\u2019t an image: {path}",
-  "twin.lookTooBig": "That screenshot weighs {size} and the cap is {cap}. Crop it or export it as JPEG: it\n  isn\u2019t shrunk automatically, because shrinking it would change what gets judged without\n  telling you.",
-  "twin.lookRejected": "Couldn\u2019t look at that ({status}). {detail}",
-  "next.title": "What\u2019s next",
-  "next.count": "{n} project{s} with something to propose",
-  "next.andMore": "and {n} more project{s} with something to propose",
-  "next.onlyNorth": "{n} more project{s}, waiting for you to say what \u201Cfinished\u201D means there.",
-  "next.onlyNorthHint": 'Say it with: panoma north <project> "\u2026"   \xB7   or on its card: {api}/p/<project>',
-  "next.nothing": "Nothing to propose: what\u2019s in the catalog is where it should be.",
-  "next.nothingHint": "That\u2019s neither praise nor a bug: it\u2019s what comes out of looking at README, health, dependencies, advisories, unsaved work and months idle.",
-  "next.tooOld": "This catalog can\u2019t say what\u2019s next yet.",
-  "next.tooOldHint": "It\u2019s older than this CLI. Update it and bring it back up: panoma down && panoma up",
-  "next.badReport": "The catalog replied with something that isn\u2019t a report.",
-  "next.httpError": "The catalog returned {status}. {detail}",
-  "next.noNorth": "Nobody has written what \u201Cfinished\u201D means here, so any ordering is guesswork.",
-  "next.noNorthAll": "Without knowing what \u201Cfinished\u201D means, everything proposed below is guesswork.",
-  "next.noNorthCount": "No north written: {n} on this list.",
-  "next.resume": "Pick it back up",
-  "next.presentable": "Make it presentable",
-  "next.plan": "A prioritized plan",
-  "next.competitors": "What it\u2019s up against",
-  "next.whyUnsaved": "{n} unsaved-work warning",
-  "next.whyUnsavedMany": "{n} unsaved-work warnings",
-  "next.whyNoReadme": "there is no README that explains it",
-  "next.whyNeverBuilt": "nobody has ever checked whether it still builds",
-  "next.whyIdle": "idle for a month",
-  "next.whyIdleMany": "idle for {n} months",
-  "next.whyAdvisories": "{n} open security advisory",
-  "next.whyAdvisoriesMany": "{n} open security advisories",
-  "next.whyOutdated": "{n} outdated direct dependency",
-  "next.whyOutdatedMany": "{n} outdated direct dependencies",
-  "next.whyLowHealth": "health {n} out of 100",
-  "next.whyLongIdle": "{n} months idle: the question is no longer about maintenance",
-  "next.whyCritiques": "{n} thing showing without opening the project",
-  "next.whyCritiquesMany": "{n} things showing without opening the project",
-  "next.noSuchProject": "\u201C{query}\u201D isn\u2019t in what\u2019s next today.",
-  "next.noSuchProjectHint": "Today\u2019s list comes out of: panoma next",
-  "next.noSuchMove": "\u201C{kind}\u201D isn\u2019t proposed on {name} today.",
-  "next.noSuchMoveHint": "What is proposed there: {kinds}",
-  "next.launching": "Opening your agent on {name}\u2026",
-  "next.launched": "working on {name}",
-  "north.title": "Each project\u2019s north: what \u201Cfinished\u201D means there",
-  "north.quote": "\u201C{north}\u201D",
-  "north.blankCount": "No north written: {n} of {total} project{s}.",
-  "north.blankHint": 'Say it in one line, looking at the project: panoma north <project> "\u2026"',
-  "north.allWritten": "{n} project{s} in the catalog, and not one is left without a north.",
-  "north.noneWritten": "No north has been written yet.",
-  "north.unlistedCount": "Norths written that can\u2019t be read from here: {n}, those of projects with nothing pending.",
-  "north.rewrite": 'Rewrite it with: panoma north {slug} "\u2026"',
-  "north.sayIt": "Say it in one line, looking at the project:",
-  "north.card": "or on its card: {url}",
-  "north.unlistedOne": "There is a north written here and this terminal can\u2019t read it.",
-  "north.unlistedOneHint": "The daily report only carries projects with something pending, and nothing is pending in this one.",
-  "north.blind": "A north that cannot be read is not replaced blind.",
-  "north.blindHint": "Whatever is written would be gone before you ever saw it. If you want to replace it anyway: --force",
-  "north.saved": "North written in {name}",
-  "north.savedOver": "North replaced in {name}",
-  "north.replaced": "It replaces this one, which the catalog no longer keeps:",
-  "north.replacedBlind": "The one that was there couldn\u2019t be read, so not even an echo of it is left.",
-  "north.rejected": "The catalog could not store the north ({status}). {detail}",
-  "north.noIdentityHint": "Scan it again and the sentence will have somewhere to live: panoma scan {root} --save",
-  "north.noSuchProject": "\u201C{query}\u201D isn\u2019t in the catalog.",
-  "north.noSuchProjectHint": "The exact names, with their north, come out of: panoma north",
-  "review.checking": "Checking the project against itself\u2026",
-  "review.title": "What is wrong, and provable without opening it",
-  "review.count": "{n} mechanical problems \xB7 {files} file{s} read",
-  "review.countOne": "1 mechanical problem \xB7 {files} file{s} read",
-  "review.groupColor": "Colours that are not in the palette",
-  "review.groupRadius": "Corners that do not match",
-  "review.groupImage": "Images that do not say what they show",
-  "review.groupLink": "Links that lead nowhere",
-  "review.row": "{place} \xB7 {claim} \u2014 {reason}",
-  "review.rowBare": "{claim} \u2014 {reason}",
-  "review.more": "and {n} more of the same",
-  "review.colorDrift": "used once or twice; the project uses {hint}",
-  "review.radiusDrift": "and {hint} are the same corner by eye, written several ways",
-  "review.imageNoAlt": "does not say what it shows: whoever cannot see it, misses it",
-  "review.linkMissing": "does not exist in the project",
-  "review.linkMoved": "missing; there is one at {path}",
-  "review.truncated": "the project has more files than the index holds: links and palette were not checked",
-  "review.next": "None of this needs judgement: fix it and run panoma review again.",
-  "review.clean": "Nothing to report: {files} file{s} read, not one mechanical problem.",
-  "review.cleanHint": "What is left \u2014whether it looks good, whether it matches the rest\u2014 needs an eye, and this command hasn\u2019t got one.",
-  "usage.next": "Usage: panoma next   \xB7   panoma next <project> <assignment>",
-  "usage.open": "Usage: panoma open <project>   \xB7   --folder   \xB7   --terminal",
-  "usage.describe": "Usage: panoma describe <project>",
-  "usage.search": 'Usage: panoma search "text to find"',
-  "usage.run": "Usage: panoma run <project> <package> [version]   \xB7   panoma run <project> --security",
-  "usage.agentKey": 'Missing the name: panoma agent-key "Claude Code"'
-};
-var CLI_LANGUAGE = "en";
-function say(key, vars = {}) {
-  return MESSAGES[key].replace(/\{(\w+)\}/g, (hole, name) => {
-    const value = vars[name];
-    return value === void 0 ? hole : String(value);
-  });
-}
-function plural(n, many = "s", one = "") {
-  return n === 1 ? one : many;
-}
-function riskText(risk) {
-  const n = risk.count ?? 0;
-  const vars = { n, s: plural(n), es: plural(n, "es") };
-  if (risk.code === "unversioned") return say("risk.unversioned");
-  if (risk.code === "untracked") return say("risk.untracked", vars);
-  if (risk.code === "no-commits") {
-    return n > 0 ? say("risk.no-commits.n", vars) : say("risk.no-commits");
-  }
-  return say(n === 1 ? `risk.${risk.code}` : `risk.${risk.code}.n`, vars);
-}
-
-export {
-  CLI_LANGUAGE,
-  say,
-  plural,
-  riskText
-};