unslop-design 1.11.0 → 1.13.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -212,6 +212,18 @@ python3 scripts/unslop-score.py examples/swiftui/after # 100/100
212
212
  The SwiftUI folders also hold the HTML mock of each state at iPhone size,
213
213
  the way quick mode shows a native before/after when there's no simulator.
214
214
 
215
+ ## Scores in the wild
216
+
217
+ Ran it on your own app or site? [Share the before/after and the
218
+ score](https://github.com/marcosnovo/unslop-design/issues/new?template=score-report.yml),
219
+ and it goes here with your permission: what the app is, the two numbers,
220
+ one image. What it missed is just as useful: that's how rules get added.
221
+
222
+ | App | Platform | Score | |
223
+ |---|---|---|---|
224
+ | [Refil](https://refil.app) (the author's) | iOS, Android | before/after above | [screens](#one-fix-at-a-time) |
225
+ | *Yours?* | | | [share it](https://github.com/marcosnovo/unslop-design/issues/new?template=score-report.yml) |
226
+
215
227
  ## Use it
216
228
 
217
229
  Type `/unslop-design` (or `$unslop-design` in Codex), or just ask in plain
@@ -288,24 +300,51 @@ edit just added, with the fix number. It says nothing when the edit is
288
300
  clean, never changes code itself, and turns off with `UNSLOP_HOOK=off` or
289
301
  `hook: off` in `.unslop/context.md`.
290
302
 
303
+ **While other agents write.** The same hook works wherever an agent's hooks
304
+ can hand a note back to the model. One command sets it up in the project:
305
+
306
+ ```bash
307
+ npx unslop-design hooks gemini --write # .gemini/settings.json, AfterTool
308
+ npx unslop-design hooks copilot --write # .github/hooks/unslop-design.json, postToolUse (preview)
309
+ npx unslop-design hooks codex # prints the ~/.codex/config.toml block, PostToolUse
310
+ ```
311
+
312
+ Each runs `npx -y unslop-design hook` after a write or edit, reads the
313
+ agent's event (Gemini's `write_file`/`replace`, Codex's `apply_patch`,
314
+ Copilot's tool arguments) and answers in the format they share. Cursor's and
315
+ Windsurf's post-edit hooks can't send anything back to the agent yet, so
316
+ there is no hook for them; the scanner and the CI work there as anywhere.
317
+
291
318
  **On every pull request (GitHub Actions).** Each hit becomes an annotation
292
- on its line, the score goes to the job summary, and the job fails under your
293
- minimum:
319
+ on its line, and one comment on the PR keeps the score and the change
320
+ against the base branch, updated on every push:
321
+
322
+ > ### unslop score: 88/100 (+7 vs main)
323
+ > **Worse** · [#68] Border and shadow on one card: 0 → 2 (-2)
324
+ > **Better** · [#32] Colored glows: 3 → 0 (+8) · [#33] Emoji in the UI: 4 → 0 (+8)
325
+
326
+ The job fails under your minimum, if you set one:
294
327
 
295
328
  ```yaml
296
329
  # .github/workflows/unslop.yml
297
330
  on: pull_request
331
+ permissions:
332
+ contents: read
333
+ pull-requests: write # for the comment
298
334
  jobs:
299
335
  unslop:
300
336
  runs-on: ubuntu-latest
301
337
  steps:
302
- - uses: actions/checkout@v4
303
- - uses: marcosnovo/unslop-design@v1.11.0
338
+ - uses: actions/checkout@v5
339
+ - uses: marcosnovo/unslop-design@v1.13.0
304
340
  with:
305
341
  paths: src
306
- min: 80 # optional; also: allow, scale, type-scale
342
+ min: 80 # optional; also: allow, scale, type-scale, comment: false
307
343
  ```
308
344
 
345
+ Locally, `unslop-score.py --json > main.json` on one branch and
346
+ `--baseline main.json` on another gives the same comparison.
347
+
309
348
  **Anywhere else.** `python3 scripts/slop-scan.py src --github` prints the
310
349
  same annotations, `--json` every hit for other tools, and `--list-rules`
311
350
  every rule id. All of them read `.unslop/context.md`, so your scale and your
package/bin/unslop.mjs CHANGED
@@ -2,6 +2,7 @@
2
2
  // npx unslop-design <command>: one entry point for the installer and the scripts.
3
3
  // No dependencies. scan and score need Python 3; audit needs Playwright.
4
4
  import { spawnSync } from "node:child_process";
5
+ import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
5
6
  import { dirname, join } from "node:path";
6
7
  import { fileURLToPath } from "node:url";
7
8
 
@@ -20,6 +21,8 @@ const HELP = `unslop-design: make your app or site stop looking AI-generated
20
21
  the rendered page at 390/768/1280 px (needs Playwright)
21
22
  npx unslop-design rules
22
23
  every rule id with its fix number
24
+ npx unslop-design hooks <gemini|codex|copilot|claude> [--write]
25
+ the post-edit hook for that agent: prints its config, --write adds it to this project
23
26
 
24
27
  Then, in your agent: "unslop this screen". Docs: https://github.com/marcosnovo/unslop-design`;
25
28
 
@@ -38,6 +41,72 @@ function python(script, args) {
38
41
  process.exit(2);
39
42
  }
40
43
 
44
+ // The same hook for every agent: `npx -y unslop-design hook` reads the agent's event on stdin.
45
+ const HOOK = "npx -y unslop-design hook";
46
+ const HOOKS = {
47
+ gemini: {
48
+ file: ".gemini/settings.json",
49
+ note: "Gemini CLI: AfterTool on write_file and replace. Project hooks ask for approval the first time.",
50
+ merge: (cfg) => {
51
+ cfg.hooks ??= {};
52
+ cfg.hooks.AfterTool = withOurHook(cfg.hooks.AfterTool,
53
+ { matcher: "write_file|replace", hooks: [{ name: "unslop-design", type: "command", command: HOOK, timeout: 30000 }] });
54
+ return cfg;
55
+ },
56
+ },
57
+ copilot: {
58
+ file: ".github/hooks/unslop-design.json",
59
+ note: "GitHub Copilot (VS Code agent, Copilot CLI, cloud agent): postToolUse. Agent hooks are in preview; commit the file for cloud sessions.",
60
+ merge: () => ({ version: 1, hooks: { postToolUse: [{ type: "command", command: HOOK, timeoutSec: 30 }] } }),
61
+ },
62
+ claude: {
63
+ file: ".claude/settings.json",
64
+ note: "Claude Code: the plugin already ships this hook (/plugin install unslop-design@unslop). Use this only without the plugin.",
65
+ merge: (cfg) => {
66
+ cfg.hooks ??= {};
67
+ cfg.hooks.PostToolUse = withOurHook(cfg.hooks.PostToolUse,
68
+ { matcher: "Write|Edit|MultiEdit", hooks: [{ type: "command", command: HOOK, timeout: 30 }] });
69
+ return cfg;
70
+ },
71
+ },
72
+ codex: {
73
+ file: null,
74
+ note: "Codex CLI: PostToolUse on apply_patch. Add this to ~/.codex/config.toml (hooks may need enabling in your version, and some versions don't fire it for apply_patch).",
75
+ text: `[[hooks.PostToolUse]]\nmatcher = "apply_patch"\n\n[[hooks.PostToolUse.hooks]]\ntype = "command"\ncommand = "${HOOK}"\n`,
76
+ },
77
+ };
78
+
79
+ // Replace only our own handler in a list of hook groups: other commands in the same group stay.
80
+ function withOurHook(groups, group) {
81
+ const ours = (h) => typeof h?.command === "string" && h.command.includes("unslop-design");
82
+ const kept = (groups || []).map(g => ({ ...g, hooks: (g.hooks || []).filter(h => !ours(h)) })).filter(g => g.hooks.length);
83
+ return [...kept, group];
84
+ }
85
+
86
+ function hooks([agent, ...flags]) {
87
+ const h = HOOKS[agent];
88
+ if (!h) {
89
+ console.error(`usage: npx unslop-design hooks <${Object.keys(HOOKS).join("|")}> [--write]\n` +
90
+ "Cursor and Windsurf hooks can't send a note back to the agent yet, so there is no hook for them.");
91
+ process.exit(2);
92
+ }
93
+ console.log(h.note + "\n");
94
+ if (!h.file) { console.log(h.text); return; }
95
+ let cfg = {};
96
+ if (existsSync(h.file)) {
97
+ try { cfg = JSON.parse(readFileSync(h.file, "utf8")); }
98
+ catch { console.error(`${h.file} isn't valid JSON: fix it or add the hook by hand.`); process.exit(2); }
99
+ }
100
+ const next = JSON.stringify(h.merge(cfg), null, 2) + "\n";
101
+ if (flags.includes("--write")) {
102
+ mkdirSync(dirname(h.file), { recursive: true });
103
+ writeFileSync(h.file, next);
104
+ console.log(`Wrote ${h.file}. The hook runs after each UI edit and is silent when it's clean.`);
105
+ } else {
106
+ console.log(`${h.file}:\n${next}\nRun again with --write to save it.`);
107
+ }
108
+ }
109
+
41
110
  switch (cmd) {
42
111
  case "install": {
43
112
  const code = run("bash", [join(ROOT, "install.sh"), ...rest]);
@@ -50,6 +119,11 @@ switch (cmd) {
50
119
  case "scan": python("slop-scan.py", rest); break;
51
120
  case "score": python("unslop-score.py", rest); break;
52
121
  case "rules": python("slop-scan.py", ["--list-rules"]); break;
122
+ case "hook": {
123
+ const r = spawnSync(process.execPath, [join(ROOT, "hooks", "unslop-hook.mjs")], { stdio: "inherit" });
124
+ process.exit(r.status ?? 0);
125
+ }
126
+ case "hooks": hooks(rest); break;
53
127
  case "audit": process.exit(run(process.execPath, [join(ROOT, "scripts", "web-audit.mjs"), ...rest]) ?? 1);
54
128
  case undefined: case "help": case "-h": case "--help": console.log(HELP); break;
55
129
  default:
@@ -1,17 +1,27 @@
1
1
  #!/usr/bin/env node
2
- // PostToolUse hook: scan the UI file the agent just wrote or edited with
2
+ // Post-edit hook: scan the UI files the agent just wrote or edited with
3
3
  // scripts/slop-scan.py and hand the new hits back to the agent as context.
4
- // Silent when clean, when the file isn't UI, when Python is missing, or when
5
- // turned off (UNSLOP_HOOK=off, or a line "hook: off" in .unslop/context.md).
4
+ //
5
+ // One script for every agent whose hooks can talk back to the model:
6
+ // Claude Code PostToolUse Write, Edit, MultiEdit: tool_input.file_path
7
+ // Gemini CLI AfterTool write_file, replace: tool_input.file_path
8
+ // Codex CLI PostToolUse apply_patch: the paths inside the patch text
9
+ // Copilot / VS Code agents PostToolUse: a file path in tool_input or toolArgs
10
+ // All of them read {"hookSpecificOutput": {"additionalContext": "…"}} on stdout.
11
+ //
12
+ // Silent when clean, when no UI file was touched, when Python is missing, or
13
+ // when turned off (UNSLOP_HOOK=off, or a line "hook: off" in .unslop/context.md).
6
14
  // It never edits anything: the agent decides, the user approves.
7
15
  import { spawnSync } from "node:child_process";
8
16
  import { existsSync, readFileSync } from "node:fs";
9
- import { dirname, join, relative, extname } from "node:path";
17
+ import { dirname, join, relative, resolve, extname } from "node:path";
10
18
  import { fileURLToPath } from "node:url";
11
19
 
12
20
  const UI = new Set([".css", ".scss", ".sass", ".less", ".html", ".htm", ".jsx", ".tsx", ".vue", ".svelte", ".astro", ".mdx", ".swift", ".kt"]);
13
21
  const SCAN = join(dirname(fileURLToPath(import.meta.url)), "..", "scripts", "slop-scan.py");
14
22
  const MAX = 8;
23
+ const PATH_KEYS = ["file_path", "filePath", "path", "absolute_path", "absolutePath", "target_file", "filename"];
24
+ const WRITE_TOOLS = /^(Write|write_file|create_file|createFile|create)$/;
15
25
 
16
26
  function quit() { process.exit(0); }
17
27
 
@@ -21,12 +31,38 @@ let ev;
21
31
  try { ev = JSON.parse(input); } catch { quit(); }
22
32
  if (/^(off|0|false|no)$/i.test(process.env.UNSLOP_HOOK || "")) quit();
23
33
 
24
- const tool = ev.tool_input || {};
25
- const file = tool.file_path;
26
- if (!file || !UI.has(extname(file).toLowerCase()) || !existsSync(file)) quit();
34
+ const cwd = ev.cwd || process.cwd();
35
+ let tool = ev.tool_input ?? ev.toolArgs ?? ev.tool_args ?? {};
36
+ if (typeof tool === "string") { try { tool = JSON.parse(tool); } catch { tool = { input: tool }; } }
37
+ const toolName = ev.tool_name || ev.toolName || "";
38
+
39
+ // The files this edit touched -> the text it brought in (null: the whole file is new).
40
+ const files = new Map();
41
+ const fresh = [tool.new_string, tool.newString, ...(Array.isArray(tool.edits) ? tool.edits : [])
42
+ .map(e => e && (e.new_string ?? e.newString))].filter(s => typeof s === "string" && s);
43
+ const whole = WRITE_TOOLS.test(toolName) || !fresh.length;
44
+ for (const key of PATH_KEYS) {
45
+ if (typeof tool[key] === "string" && tool[key]) files.set(resolve(cwd, tool[key]), whole ? null : fresh.join("\n"));
46
+ }
47
+ // Codex apply_patch: "*** Update File: src/Card.tsx" then "+added line" rows.
48
+ const strings = [];
49
+ (function walk(v) {
50
+ if (typeof v === "string") strings.push(v);
51
+ else if (Array.isArray(v)) v.forEach(walk);
52
+ else if (v && typeof v === "object") Object.values(v).forEach(walk);
53
+ })(tool);
54
+ for (const s of strings) {
55
+ for (const block of s.split(/^(?=\*\*\* (?:Add|Update|Delete) File: )/m)) {
56
+ const m = block.match(/^\*\*\* (Add|Update) File: (.+)$/m);
57
+ if (!m) continue;
58
+ const added = block.split("\n").filter(l => l.startsWith("+") && !l.startsWith("+++")).map(l => l.slice(1));
59
+ files.set(resolve(cwd, m[2].trim()), m[1] === "Add" ? null : added.join("\n"));
60
+ }
61
+ }
62
+ const targets = [...files].filter(([f]) => UI.has(extname(f).toLowerCase()) && existsSync(f));
63
+ if (!targets.length) quit();
27
64
 
28
65
  // Project context (written by "init"): slop-scan reads its scales and allowed rules itself.
29
- const cwd = ev.cwd || process.cwd();
30
66
  const opts = [];
31
67
  const ctxPath = join(cwd, ".unslop", "context.md");
32
68
  if (existsSync(ctxPath)) {
@@ -35,8 +71,8 @@ if (existsSync(ctxPath)) {
35
71
  }
36
72
 
37
73
  let out = null;
38
- for (const py of ["python3", "python"]) {
39
- const r = spawnSync(py, [SCAN, file, "--json", ...opts], { encoding: "utf8", timeout: 15000 });
74
+ for (const py of process.platform === "win32" ? ["py", "python", "python3"] : ["python3", "python"]) {
75
+ const r = spawnSync(py, [SCAN, ...targets.map(([f]) => f), "--json", ...opts], { encoding: "utf8", timeout: 15000 });
40
76
  if (r.error) continue; // not installed under this name
41
77
  if (r.status !== 0) quit(); // a bad --allow in context.md, etc.: stay out of the way
42
78
  out = r.stdout;
@@ -46,29 +82,34 @@ if (!out) quit();
46
82
 
47
83
  let hits;
48
84
  try { hits = JSON.parse(out).hits; } catch { quit(); }
49
- // After an Edit, only what the edit brought in: hits on the lines its new text now spans
50
- // (a partial-line edit counts its whole line).
51
- const fresh = [tool.new_string, ...(tool.edits || []).map(e => e.new_string)].filter(Boolean);
52
- if (fresh.length && ev.tool_name !== "Write") {
53
- const text = readFileSync(file, "utf8");
54
- const touched = new Set();
55
- for (const ns of fresh) {
56
- for (let at = text.indexOf(ns); at !== -1; at = text.indexOf(ns, at + Math.max(1, ns.length))) {
57
- const first = text.slice(0, at).split("\n").length;
58
- const last = first + ns.split("\n").length - 1;
59
- for (let n = first; n <= last; n++) touched.add(n);
85
+
86
+ // After an edit, only the lines its new text now sits on (a partial-line edit counts its whole line).
87
+ const touched = new Map();
88
+ for (const [f, text] of targets) {
89
+ if (text === null) continue;
90
+ const body = readFileSync(f, "utf8");
91
+ const lines = new Set();
92
+ // Every non-empty piece counts, even "3": a short value is a common UI edit. A short piece may
93
+ // also match other lines; reporting a few extra beats missing what the edit added.
94
+ for (const piece of text.split("\n").map(l => l.trim()).filter(Boolean)) {
95
+ for (let at = body.indexOf(piece); at !== -1; at = body.indexOf(piece, at + piece.length)) {
96
+ lines.add(body.slice(0, at).split("\n").length);
60
97
  }
61
98
  }
62
- hits = hits.filter(h => touched.has(h.line));
99
+ touched.set(f, lines);
63
100
  }
101
+ hits = hits.filter(h => {
102
+ const lines = touched.get(resolve(h.file));
103
+ return !lines || lines.has(h.line);
104
+ });
64
105
  if (!hits.length) quit();
65
- hits.sort((a, b) => a.line - b.line);
106
+ hits.sort((a, b) => (a.file === b.file ? a.line - b.line : a.file < b.file ? -1 : 1));
66
107
 
67
- const rel = relative(cwd, file) || file;
68
- const shown = hits.slice(0, MAX).map(h => `- line ${h.line}: ${h.ref} ${h.rule} — ${h.description}`);
69
- const more = hits.length > MAX ? `\n- … ${hits.length - MAX} more (python3 scripts/slop-scan.py ${rel})` : "";
70
- const note = `unslop-design found ${hits.length} slop pattern${hits.length === 1 ? "" : "s"} in what you just wrote to ${rel}:\n` +
108
+ const shown = hits.slice(0, MAX).map(h => `- ${relative(cwd, resolve(h.file)) || h.file}:${h.line}: ${h.ref} ${h.rule} — ${h.description}`);
109
+ const more = hits.length > MAX ? `\n- … ${hits.length - MAX} more (run scripts/slop-scan.py on the same files)` : "";
110
+ const note = `unslop-design found ${hits.length} slop pattern${hits.length === 1 ? "" : "s"} in what you just wrote:\n` +
71
111
  shown.join("\n") + more +
72
112
  "\nFix them by changing the value or the token (checklist.md has each fix), not by hiding them. " +
73
113
  "If one is a written project decision, it belongs in .unslop/context.md under allow:, not in an unslop-ignore.";
74
- process.stdout.write(JSON.stringify({ hookSpecificOutput: { hookEventName: "PostToolUse", additionalContext: note } }));
114
+ const event = ev.hook_event_name || ev.hookEventName || "PostToolUse";
115
+ process.stdout.write(JSON.stringify({ hookSpecificOutput: { hookEventName: event, additionalContext: note } }));
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "unslop-design",
3
- "version": "1.11.0",
3
+ "version": "1.13.0",
4
4
  "description": "Make your app or site stop looking AI-generated: install the agent skill, scan UI code for slop and score it 0-100.",
5
5
  "license": "MIT",
6
6
  "author": "Marcos Novo (https://github.com/marcosnovo)",
@@ -3,7 +3,7 @@
3
3
 
4
4
  python3 scripts/unslop-score.py [paths…] [--audit unslop-audit/report.json]
5
5
  [--scale …] [--json | --markdown] [--min 80]
6
- [--save [.unslop]]
6
+ [--save [.unslop]] [--baseline base.json]
7
7
 
8
8
  Scores the source with the same rules as `slop-scan.py` and, if you pass a
9
9
  `web-audit.mjs` report, the rendered page too (scroll, contrast, focus,
@@ -13,7 +13,9 @@ are left more than how big the codebase is. Prints the breakdown and a
13
13
  README badge. With --min it exits 1 below that score (pre-commit hook).
14
14
  With --save it appends the run to .unslop/history.json and prints the
15
15
  trend ("score 41 → 88 → 91"); --markdown prints a table for a PR comment or
16
- a GitHub Actions job summary.
16
+ a GitHub Actions job summary. --baseline compares with an earlier --json
17
+ run (the base branch in a PR): "score 72 → 81 (+9)", what got better and what
18
+ got worse.
17
19
 
18
20
  It is a smoke alarm, not a judge: a 95 can still have a flat hierarchy,
19
21
  and a gradient you chose on purpose costs points anyway (silence that line
@@ -179,6 +181,26 @@ def save_run(folder, run):
179
181
  return [h["score"] for h in history]
180
182
 
181
183
 
184
+ def compare(base, score, lines):
185
+ """Score delta and per-row changes against an earlier --json output."""
186
+ before = {r["rule"]: (r["count"], r["points"]) for r in base.get("penalties", [])}
187
+ now = {r: (n, p) for p, r, n in lines}
188
+ better, worse = [], []
189
+ for rule in sorted(set(before) | set(now)):
190
+ (bn, bp), (nn, np_) = before.get(rule, (0, 0)), now.get(rule, (0, 0))
191
+ if np_ < bp:
192
+ better.append({"rule": rule, "from": bn, "to": nn, "points": round(bp - np_, 1)})
193
+ elif np_ > bp:
194
+ worse.append({"rule": rule, "from": bn, "to": nn, "points": round(np_ - bp, 1)})
195
+ better.sort(key=lambda x: -x["points"])
196
+ worse.sort(key=lambda x: -x["points"])
197
+ return {"score": base["score"], "delta": score - base["score"], "better": better, "worse": worse}
198
+
199
+
200
+ def signed(n):
201
+ return f"+{n:g}" if n > 0 else f"{n:g}" if n < 0 else "±0"
202
+
203
+
182
204
  def main():
183
205
  ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
184
206
  ap.add_argument("paths", nargs="*", help="source to scan (default: . unless only --audit is given)")
@@ -189,6 +211,8 @@ def main():
189
211
  ap.add_argument("--save", nargs="?", const=".unslop", metavar="DIR",
190
212
  help="append this run to DIR/history.json (default .unslop) and show the trend")
191
213
  ap.add_argument("--min", type=float, help="exit 1 below this score")
214
+ ap.add_argument("--baseline", metavar="JSON", help="an earlier --json output to compare with (e.g. the base branch)")
215
+ ap.add_argument("--baseline-label", default="baseline", metavar="NAME", help="what to call it (e.g. main)")
192
216
  ap.add_argument("--type-scale", help="allowed font sizes in px (default: the project context, else slop-scan's)")
193
217
  ap.add_argument("--context", metavar="FILE", help="project context to read (default .unslop/context.md if present)")
194
218
  ap.add_argument("--no-context", action="store_true", help="ignore .unslop/context.md")
@@ -255,12 +279,34 @@ def main():
255
279
  scores = save_run(a.save, {"score": score, "files": files, "audit": bool(a.audit), "allowed": allowed,
256
280
  "penalties": [{"rule": r, "count": n, "points": p} for p, r, n in lines]})
257
281
  trend = " → ".join(str(s) for s in scores[-5:])
282
+ cmp_ = None
283
+ if a.baseline:
284
+ try:
285
+ with open(a.baseline, encoding="utf-8") as fh:
286
+ cmp_ = compare(json.load(fh), score, lines)
287
+ except (OSError, ValueError, KeyError) as e:
288
+ sys.exit(f"can't read the baseline {a.baseline}: {e}. It is the --json output of an earlier run.")
258
289
  if a.markdown:
259
- out = [f"### unslop score: {score}/100 ({word})", ""]
290
+ if cmp_:
291
+ out = [f"### unslop score: {score}/100 ({signed(cmp_['delta'])} vs {a.baseline_label})", "",
292
+ f"{a.baseline_label} {cmp_['score']} → this change {score}", ""]
293
+ for title, rows, sign in (("Worse", cmp_["worse"], "-"), ("Better", cmp_["better"], "+")):
294
+ if rows:
295
+ out.append(f"**{title}**")
296
+ out += [f"- {r['rule']}: {r['from']} → {r['to']} ({sign}{r['points']:g})" for r in rows]
297
+ out.append("")
298
+ if not cmp_["better"] and not cmp_["worse"]:
299
+ out += ["No change in any row.", ""]
300
+ if lines:
301
+ out += ["<details><summary>Every row now</summary>", ""]
302
+ else:
303
+ out = [f"### unslop score: {score}/100 ({word})", ""]
260
304
  if trend and "→" in trend:
261
305
  out += [f"Trend: {trend}", ""]
262
306
  if lines:
263
307
  out += ["| Points | Fix | Hits |", "|---:|---|---:|"] + [f"| -{p:g} | {r} | {n} |" for p, r, n in lines]
308
+ if cmp_:
309
+ out += ["", "</details>"]
264
310
  else:
265
311
  out.append("Nothing found.")
266
312
  if allowed:
@@ -270,7 +316,7 @@ def main():
270
316
  elif a.json:
271
317
  print(json.dumps({"score": score, "verdict": word, "files": files, "audit": bool(a.audit), "badge": badge,
272
318
  "penalties": [{"rule": r, "count": n, "points": p} for p, r, n in lines],
273
- "allowed": allowed, "trend": trend}, indent=2))
319
+ "allowed": allowed, "trend": trend, "baseline": cmp_}, indent=2))
274
320
  else:
275
321
  src = []
276
322
  if paths:
@@ -282,6 +328,11 @@ def main():
282
328
  print(f"unslop score: {score}/100 — {word} ({' + '.join(src)})")
283
329
  if trend and "→" in trend:
284
330
  print(f"trend: {trend} ({os.path.join(a.save, 'history.json')})")
331
+ if cmp_:
332
+ print(f"vs {a.baseline_label}: {cmp_['score']} → {score} ({signed(cmp_['delta'])})")
333
+ for title, rows, sign in (("worse", cmp_["worse"], "-"), ("better", cmp_["better"], "+")):
334
+ for r in rows:
335
+ print(f" {title}: {r['rule']} {r['from']} → {r['to']} ({sign}{r['points']:g})")
285
336
  if allowed:
286
337
  print("allowed by the project (--allow, not scored): " + ", ".join(allowed))
287
338
  if lines: