@xsft/panels 0.5.1 → 0.6.0

package/README.md

@@ -31,7 +31,7 @@ Run the command in the folder of the product. Each judge starts in that folder,
 
 On a terminal, `panels run` shows the score table, and each judge runs in its interactive harness in a pseudo-terminal inside the command. Each judge writes `judges/N/answer.json` and stays open. The table writes the report when each judge has an answer or an error, and the command exits from `report.json` when you quit. Without a terminal, the judges run in the background (`claude -p`, `codex exec`).
 
-A tab strip on the last three lines shows a hairline, a tab per view, and a status line with the progress, the elapsed time, and the keys. The tabs are the scores, then each judge with its number, name, and mark (score, ✗ for an error, · while queued, ! while the harness asks whether to trust the folder). The label of a running judge glimmers. The strip is the navigation area. With the focus on the strip, `←` and `→` move the active tab and switch the view, `1` to `9` jump to a started judge, `0` to the scores, `enter` gives the focus to the judge, and `q` quits and asks `y` while a judge still runs. With the focus on a judge, each key goes to the judge, and `ctrl-q` gives the focus back to the strip. The scores take no input, so the strip keeps the focus there. A harness that has not seen the folder asks whether to trust it. The first judge starts alone; press `1`, answer, and the other judges start.
+A tab strip on the last three lines shows a hairline, a tab per view, and a status line with the progress, the elapsed time, and the keys. The tabs are the scores, then each judge with its number, name, and mark (score, ✗ for an error, · while queued, ! while the harness asks whether to trust the folder). One soft wave sweeps across the strip and lights the labels of the running judges as it passes; the same wave sweeps the running rows of the scores table. The strip is the navigation area. With the focus on the strip, `←` and `→` move the active tab and switch the view, `1` to `9` jump to a started judge, `0` to the scores, `enter` gives the focus to the judge, and `q` quits; while a judge still runs, the first `q` asks and a second `q` quits. With the focus on a judge, each key goes to the judge, and `ctrl-q` gives the focus back to the strip. The scores take no input, so the strip keeps the focus there. A harness that has not seen the folder asks whether to trust it. The first judge starts alone; press `1`, answer, and the other judges start.
 
 ## Run folder
 

package/lib/judge.js

@@ -9,8 +9,6 @@ const TOOLS = 'Read,Glob,Grep,Bash,WebFetch,WebSearch';
 // With --permission-mode dontAsk, Claude refuses every tool call that this list does not allow, so Bash runs git reads only.
 const ALLOWED_TOOLS = 'Read,Glob,Grep,WebFetch,WebSearch,Bash(git log:*),Bash(git show:*),Bash(git diff:*),Bash(git status:*),Bash(git ls-files:*),Bash(git rev-parse:*),Bash(git branch --list:*)';
 
-export const bandToScore = (band) => (band <= 3 ? 1 : band <= 6 ? 2 : band <= 8 ? 3 : 4);
-
 /**
  * The user message of a judge. criteria items are { rank, name, description, body }. Each part sits
  * in a tag, because a body is markdown with headings of its own.
@@ -45,8 +43,7 @@ export function check(answer, criteria) {
     if (missing.length) return [`Item ${i + 1} lacks ${missing.join(', ')}.`];
     const errors = [];
     if (r.rank !== c.rank || r.criterion !== c.name) errors.push(`Item ${i + 1} must have rank ${c.rank} and criterion "${c.name}".`);
-    if (!Number.isInteger(r.band) || r.band < 1 || r.band > 10) errors.push(`Rank ${c.rank}: band must be a whole number from 1 to 10.`);
-    else if (r.score !== bandToScore(r.band)) errors.push(`Rank ${c.rank}: band ${r.band} maps to score ${bandToScore(r.band)}, not ${r.score}.`);
+    if (!Number.isInteger(r.score) || r.score < 1 || r.score > 4) errors.push(`Rank ${c.rank}: score must be a whole number from 1 to 4.`);
     if (r.score < 4 && !(Array.isArray(r.fixes) && r.fixes.length)) errors.push(`Rank ${c.rank}: score ${r.score} is below 4, so fixes must list each fix that score 4 needs.`);
     return errors;
   });

package/lib/report.js

@@ -15,6 +15,8 @@ export function buildReport({ panel, target, cwd, harness, judges, outcomes }) {
   return { status, panel: { owner: panel.owner, name: panel.name, version: panel.version }, target, cwd, harness, size: judges.length, criteria, errors };
 }
 
+const LEVELS = { 1: 'Fails', 2: 'Falls short', 3: 'Almost met', 4: 'Met' };
+
 const cell = (text) => text.replaceAll('|', '\\|').replaceAll('\n', ' ');
 
 export function renderMarkdown(report) {
@@ -32,7 +34,7 @@ export function renderMarkdown(report) {
   const list = (title, items) => items.length && lines.push(title, ...items.map((item) => `- ${item}`));
   for (const c of criteria) {
     for (const { result } of c.judges) {
-      lines.push('', `## ${c.rank}. ${c.name}: ${result.score} (band ${result.band})`);
+      lines.push('', `## ${c.rank}. ${c.name}: ${result.score}/4 ${LEVELS[result.score]}`);
       list('Evidence', result.evidence.map((e) => `${e.where}: ${e.quote ?? e.claim}`));
       list('Fixes', result.fixes.map((f) => `${f.problem} ${f.change} (${f.where})`));
       list('Not checked', result.notChecked.map((n) => `${n.test}: ${n.reason}`));

package/lib/table.js

@@ -31,16 +31,21 @@ const TABLE_MODES = '\x1b[?1049h\x1b[r\x1b[0m\x1b[4l\x1b[?6l\x1b[?7h\x1b[?1l\x1b
  * trust is true while the harness of a running judge asks whether to trust the folder.
  * final is null until the report exists, then { status, path }. confirm is true while q waits for y.
  */
-// The glimmer on the text of a running judge: grey text, and a bright highlight three characters wide that
-// sweeps from left to right one character per frame and wraps.
-const GLIMMER = { base: '\x1b[38;5;244m', mid: '\x1b[38;5;250m', high: '\x1b[38;5;255m' };
-const GLIMMER_ORANGE = { base: ORANGE, mid: '\x1b[38;2;255;170;140m', high: '\x1b[38;5;255m' };
+// The glimmer on the text of the running judges: grey text, and one soft wave three columns wide that moves one
+// column per frame from the left edge of the screen to the right edge and starts over. A running label lights
+// up only where the wave passes over its own columns, so every running label lights at the same columns.
+const GLIMMER = { base: '\x1b[38;5;244m', mid: '\x1b[38;5;247m', high: '\x1b[38;5;250m' };
+const GLIMMER_ORANGE = { base: ORANGE, mid: '\x1b[38;2;255;140;100m', high: '\x1b[38;2;255;175;145m' };
+const WAVE = 3;
 
-/** text with the glimmer at frame. colours is { base, mid, high }; prefix (such as bold) goes before each piece. */
-export function glimmer(text, frame, colours = GLIMMER, prefix = '') {
-  const centre = frame % text.length;
+/**
+ * text with the glimmer at frame, where the text starts at the screen column start (0-based) on a screen width
+ * columns wide. colours is { base, mid, high }; prefix (such as bold) goes before each piece.
+ */
+export function glimmer(text, frame, start, width, colours = GLIMMER, prefix = '') {
+  const centre = (frame % (width + WAVE)) - 1;
   return [...text].map((c, i) => {
-    const d = Math.min(Math.abs(i - centre), text.length - Math.abs(i - centre));
+    const d = Math.abs(start + i - centre);
     return `${prefix}${d === 0 ? colours.high : d === 1 ? colours.mid : colours.base}${c}`;
   }).join('') + RESET;
 }
@@ -54,17 +59,17 @@ export function render({ title, rows, final, confirm, width, frame = 0 }) {
   for (const r of rows) {
     const head = `${String(r.rank).padStart(3)}  ${r.name.padEnd(pad)}`;
     if (r.state === 'running' && r.trust) lines.push(`${ORANGE}${head}press ${r.rank} and answer the trust prompt${RESET}`);
-    else if (r.state === 'running') lines.push(`${String(r.rank).padStart(3)}  ${glimmer(r.name, frame)}`);
+    else if (r.state === 'running') lines.push(`${String(r.rank).padStart(3)}  ${glimmer(r.name, frame, 5, width)}`);
     else if (r.state === 'queued') lines.push(`${head}${GREY}queued${RESET}`);
     else if (r.state === 'error') lines.push(`${head}error`);
-    else lines.push(`${head}${r.result.score} ${GREY}b${r.result.band}${RESET}`);
+    else lines.push(`${head}${r.result.score}/4`);
   }
   lines.push(rule);
   if (final) {
     const at4 = rows.filter((r) => r.state === 'done' && r.result.score === 4).length;
     lines.push(` ${BOLD}${final.status.toUpperCase()}${RESET}  ${at4} of ${rows.length} at 4`, ` report ${final.path}`);
   }
-  if (confirm) lines.push(' judges still run. y quits, another key stays');
+  if (confirm) lines.push(' judges still run. q again quits, another key stays');
   return lines;
 }
 
@@ -81,7 +86,7 @@ export function renderStrip({ rows, view, final, frame = 0, width, focus = 'stri
     if (r.state === 'running') return r.trust ? '!' : '';
     if (r.state === 'queued') return '·';
     if (r.state === 'error') return '✗';
-    return String(r.result.score);
+    return `${r.result.score}/4`;
   };
   const finished = rows.filter((r) => r.state === 'done' || r.state === 'error').length;
   const clock = `${Math.floor(elapsed / 60)}:${String(elapsed % 60).padStart(2, '0')}`;
@@ -94,14 +99,17 @@ export function renderStrip({ rows, view, final, frame = 0, width, focus = 'stri
   const used = (list) => 1 + list.reduce((n, t) => n + plain(t).length, 0) + 2 * (list.length - 1);
   const list = [0, 1, 2].map(tabs).find((l) => used(l) <= width) ?? tabs(2);
   let line = ' ';
+  let col = 1;
   list.forEach((t, i) => {
-    if (i) line += '  ';
+    if (i) { line += '  '; col += 2; }
     const active = i === view;
     const base = active ? (focus === 'strip' ? `${ORANGE}${BOLD}` : BOLD) : GREY;
     const text = plain(t);
     const row = rows[i - 1];
+    const start = col;
+    col += text.length;
     if (row?.state === 'running' && !row.trust) {
-      line += glimmer(text, frame, active && focus === 'strip' ? GLIMMER_ORANGE : GLIMMER, active ? BOLD : '');
+      line += glimmer(text, frame, start, width, active && focus === 'strip' ? GLIMMER_ORANGE : GLIMMER, active ? BOLD : '');
       return;
     }
     const pieces = [];
@@ -120,13 +128,13 @@ export function renderStrip({ rows, view, final, frame = 0, width, focus = 'stri
  * view is 0 for the scores, or N for judge N. open holds the ranks of the judges that have started. With the
  * focus on a judge, every key but ctrl-q goes to the judge, and ctrl-q moves the focus to the strip. With the
  * focus on the strip, the arrows move the active tab and switch the view, 1-9 jump to a started judge, 0 to the
- * scores, Enter, Esc, or ctrl-q give the focus back to the judge, and q quits (y confirms while a judge runs).
+ * scores, Enter, Esc, or ctrl-q give the focus back to the judge, and q quits (a second q confirms while a judge runs).
  * The scores take no input, so the strip keeps the focus there.
  */
 export function press({ view, confirm, focus }, key, { open, running }) {
   const strip = (v) => ({ view: v, confirm: false, focus: 'strip' });
   if (focus === 'judge') return TOGGLE.includes(key) ? strip(view) : { view, confirm, focus, send: true };
-  if (confirm) return { ...strip(0), quit: key === 'y' };
+  if (confirm) return { ...strip(0), quit: key === 'q' };
   if (key === 'q') return running ? { view: 0, confirm: true, focus: 'strip' } : { ...strip(0), quit: true };
   if (RIGHT.includes(key) || LEFT.includes(key)) {
     const cycle = [0, ...open];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xsft/panels",
-  "version": "0.5.1",
+  "version": "0.6.0",
   "private": false,
   "description": "Run a panel of AI judges from cross.software/panels against a target.",
   "type": "module",

package/prompts/judge-system-prompt.txt

@@ -20,22 +20,16 @@ Score your criteria one at a time, in rank order. Before you read the brief or t
 
 HOW TO SCORE
 
-1. Follow the body to a level from 1 to 10. The body calls this level a score. Here it is the band.
-2. Apply each cap of the body, such as "cannot support a score above 8". When two caps apply, use the lower cap.
-3. When the evidence fits two bands, use the lower band.
-4. Convert the band to the score with this map:
-   band 1 to 3  = score 1
-   band 4 to 6  = score 2
-   band 7 or 8  = score 3
-   band 9 or 10 = score 4
-5. A criterion passes only at score 4.
+1. Follow the body to a level from 1 to 4: 4 Met, 3 Almost met, 2 Falls short, 1 Fails. The level is the score.
+2. Apply each cap of the body, such as "cannot support a score above 3". When two caps apply, use the lower cap.
+3. When the evidence fits two levels, use the lower level.
+4. A criterion passes only at score 4.
 
 EVIDENCE
 
 - Each claim needs evidence that the engineer can check: a URL, a file and line, a quote, or a command and its output.
 - Give no credit for intent. A promise in a comment, a doc, or the brief is not done work.
-- No evidence for a band means the lower band.
-- At band 9, name the one specific gap of the body in the evidence.
+- No evidence for a level means the lower level.
 
 RECORDS AND FIXES
 
@@ -44,7 +38,7 @@ RECORDS AND FIXES
 
 PROHIBITIONS
 
-- Do not change the body, the bands, the caps, or the map.
+- Do not change the body, the levels, or the caps.
 - Do not score a fault that the body gives to another criterion.
 - Do not write or change files. Name each fix, and do not make it.
 - Do not submit forms, send messages, or post data.
@@ -58,7 +52,6 @@ Return one JSON object and nothing else. Give one item in results for each of yo
   "results": [{
     "rank": integer,
     "criterion": string,
-    "band": integer 1 to 10,
     "score": integer 1 to 4,
     "verdict": string, one sentence of 30 words or less,
     "records": [{ "name": string, "columns": [string], "rows": [[string]] }],
@@ -73,7 +66,6 @@ Example of one item:
 {
   "rank": 5,
   "criterion": "Original design assets",
-  "band": 5,
   "score": 2,
   "verdict": "The hero shows the real app, and the six feature icons are Lucide icons.",
   "records": [{

package/schema/report.json

@@ -61,11 +61,10 @@
     "judgeResult": {
       "type": "object",
       "additionalProperties": false,
-      "required": ["rank", "criterion", "band", "score", "verdict", "records", "evidence", "notChecked", "fixes"],
+      "required": ["rank", "criterion", "score", "verdict", "records", "evidence", "notChecked", "fixes"],
       "properties": {
         "rank": { "type": "integer" },
         "criterion": { "type": "string" },
-        "band": { "type": "integer", "minimum": 1, "maximum": 10 },
         "score": { "type": "integer", "minimum": 1, "maximum": 4 },
         "verdict": { "type": "string" },
         "records": {