@ibalzam/codejitsu-core 0.11.1 → 0.12.0

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/bin/codejitsu.mjs CHANGED
@@ -4,6 +4,7 @@ import { runBlog } from '../modules/cli/src/blog.mjs';
4
4
  import { runDeploySetup, runDeployTrigger } from '../modules/cli/src/deploy.mjs';
5
5
  import { runDoctor } from '../modules/cli/src/doctor.mjs';
6
6
  import { runBlogInit } from '../modules/cli/src/blog-init.mjs';
7
+ import { runBlogSelftest } from '../modules/cli/src/blog-selftest.mjs';
7
8
  import { runAudit } from '../modules/audit/src/run.mjs';
8
9
 
9
10
  const subcommand = process.argv[2];
@@ -13,6 +14,17 @@ const COMMANDS = {
13
14
  'blog:list': () => runBlog('blog:list'),
14
15
  'blog:drafts': () => runBlog('blog:drafts'),
15
16
  'blog:init': () => runBlogInit(),
17
+ 'blog:selftest': () => {
18
+ const { values } = parseArgs({
19
+ args: rest,
20
+ options: {
21
+ topic: { type: 'string' },
22
+ model: { type: 'string' },
23
+ },
24
+ allowPositionals: true,
25
+ });
26
+ return runBlogSelftest({ topic: values.topic, model: values.model });
27
+ },
16
28
  'deploy:setup': () => runDeploySetup(),
17
29
  'deploy:run': () => runDeployTrigger(),
18
30
  doctor: () => runDoctor(),
@@ -59,6 +71,7 @@ function printHelp() {
59
71
  console.log(` blog:list List every non-draft post with URL + image check`);
60
72
  console.log(` blog:drafts List future-dated (pending) posts only`);
61
73
  console.log(` blog:init Install /blog, /blog-batch, /blog-images slash commands`);
74
+ console.log(` blog:selftest Cold-Claude write a throwaway post + grade it. Flags: --topic, --model`);
62
75
  console.log(``);
63
76
  console.log(` deploy:setup Wire up daily Cloudflare deploy (prompts for hook URL)`);
64
77
  console.log(` deploy:run Trigger the Daily Deploy workflow once now`);
@@ -31,10 +31,17 @@ export async function runBlogQuality(ctx) {
31
31
  const staleCutoff = new Date(today);
32
32
  staleCutoff.setUTCMonth(staleCutoff.getUTCMonth() - staleMonths);
33
33
 
34
+ // Tag governance: approvedTags is the controlled vocabulary (cap 12).
35
+ const blogWriter = config.blogWriter && typeof config.blogWriter === 'object' ? config.blogWriter : null;
36
+ const approvedTags = blogWriter?.approvedTags ?? null;
37
+ const approvedSet = approvedTags ? new Set(approvedTags) : null;
38
+
34
39
  const stalePosts = [];
35
40
  const missingDescription = [];
36
41
  const missingImage = [];
37
42
  const shortBody = [];
43
+ const htmlBodyPosts = [];
44
+ const unapprovedTagPosts = [];
38
45
  const dupTitle = new Map();
39
46
  const dupDesc = new Map();
40
47
 
@@ -44,6 +51,22 @@ export async function runBlogQuality(ctx) {
44
51
  if (draftField && data[draftField]) continue;
45
52
 
46
53
  const slug = fileName.replace(/\.md$/, '');
54
+
55
+ // Tag governance: collect category + tags, flag any not in approvedTags.
56
+ if (approvedSet) {
57
+ const postTags = [];
58
+ if (typeof data.category === 'string') postTags.push(data.category);
59
+ if (Array.isArray(data.tags)) postTags.push(...data.tags);
60
+ const bad = postTags.filter((t) => !approvedSet.has(t));
61
+ if (bad.length > 0) {
62
+ unapprovedTagPosts.push(`${slug}: ${[...new Set(bad)].join(', ')}`);
63
+ }
64
+ }
65
+
66
+ // Body format: flag posts whose body is raw HTML (legacy WordPress imports).
67
+ if (/^\s*<(p|h[1-6]|div|ul|ol|table)[\s>]/i.test(content.trimStart())) {
68
+ htmlBodyPosts.push(slug);
69
+ }
47
70
  const dateVal = data[dateField];
48
71
  const postDate = dateVal instanceof Date ? dateVal : (typeof dateVal === 'string' ? new Date(dateVal) : null);
49
72
 
@@ -69,6 +92,30 @@ export async function runBlogQuality(ctx) {
69
92
  }
70
93
 
71
94
  results.push(info(`${files.length} blog posts indexed`));
95
+
96
+ // Tag governance results.
97
+ if (!approvedTags) {
98
+ results.push(info('No blogWriter.approvedTags configured — tag governance not enforced'));
99
+ } else {
100
+ results.push(
101
+ approvedTags.length <= 12
102
+ ? pass(`approvedTags within cap (${approvedTags.length}/12)`)
103
+ : fail(`approvedTags exceeds cap (${approvedTags.length}/12)`,
104
+ 'Trim to 12 or fewer. The cap keeps the taxonomy tight.')
105
+ );
106
+ results.push(summarize(
107
+ 'All post tags + categories are in approvedTags',
108
+ unapprovedTagPosts,
109
+ 'fail'
110
+ ));
111
+ }
112
+
113
+ results.push(summarize(
114
+ 'Post bodies are markdown (not legacy HTML)',
115
+ htmlBodyPosts,
116
+ 'warn'
117
+ ));
118
+
72
119
  results.push(summarize(
73
120
  `No posts older than ${staleMonths} months`,
74
121
  stalePosts,
@@ -143,9 +143,22 @@ say stop: leave the file and exit. They can run `/blog` row by row whenever.
143
143
 
144
144
  ## Verify
145
145
 
146
- - [ ] N rows, all unique slugs
147
- - [ ] Dates respect cadence; no collisions with existing posts
148
- - [ ] Tags balanced across `approvedTags`
149
- - [ ] Service + city distribution covers most of the config lists
150
- - [ ] Season fits each `pubDate`'s month
151
- - [ ] No duplicate or near-duplicate topics
146
+ After generating the schedule, **actively count** these distributions and
147
+ rebalance the draft if any fail. Don't just eyeball.
148
+
149
+ - N rows, all unique slugs (grep the slug column for duplicates)
150
+ - Dates respect cadence; no collisions with existing post `pubDate`s
151
+ - **Tag distribution**: no single primary tag exceeds 40% of the batch.
152
+ (For N=20, no tag appears as primary more than 8 times.)
153
+ - **Topic-format distribution**: count comparison / how-to / troubleshooting /
154
+ seasonal across the batch. No single format exceeds ~40%. Roughly
155
+ 30/30/20/20 is the target.
156
+ - **City coverage**: cover most of `blogWriter.locations` before any city
157
+ repeats. For N ≤ locations.length, every city should be distinct.
158
+ - **Service coverage**: spread across `blogWriter.services`; don't anchor half
159
+ the batch to one service.
160
+ - Season fits each `pubDate`'s month per `seasonalRules`
161
+ - No duplicate or near-duplicate topics (compare titles for keyword overlap)
162
+
163
+ If any distribution check fails, **revise the batch before writing the file** —
164
+ swap the over-represented topics for under-covered service/city/format combos.
@@ -36,7 +36,10 @@ Everything in there is **site-specific input** for what you're about to write:
36
36
  - `audience` — primary reader (REQUIRED)
37
37
  - `services` — names that map to `/services/<slug>/` (REQUIRED)
38
38
  - `locations` — names that map to `/service-areas/<slug>/` (REQUIRED)
39
- - `approvedTags` — exhaustive. If unset, derive from `blog.categories` in config OR ask the user once
39
+ - `approvedTags` — **REQUIRED for writing; the single controlled vocabulary**
40
+ (hard cap: 12 per site). Governs BOTH `category` and `tags` fields. If unset,
41
+ STOP and ask the user to define up to 12 (or derive from the `category` enum
42
+ in `src/content.config.ts`). Never write a post with a tag outside this list.
40
43
  - `wordCount` — `{ min, max, default }`. Kit default `{ 1200, 2500, 1800 }`
41
44
  - `faqs` — `{ min, max }`. Kit default `{ 5, 8 }`
42
45
  - `internalLinks` — `{ min, max }` per post. Kit default `{ 3, 6 }`
@@ -47,7 +50,28 @@ Everything in there is **site-specific input** for what you're about to write:
47
50
  - `cadenceDays` — for /blog-batch. Kit default `4`
48
51
 
49
52
  Also detect the blog's frontmatter shape by reading `src/content.config.ts` and
50
- ONE existing post in `src/content/blog/`. Use that exact shape for the new post.
53
+ ONE existing post in `src/content/blog/`. **Match the frontmatter field SHAPE
54
+ (which fields exist), but NOT the body format and NOT the field VALUES.**
55
+
56
+ **Body format: always write clean markdown.** Use `## H2`, `- bullets`,
57
+ `1. numbered`, `| markdown | tables |`, `**bold**`. Do this EVEN IF existing
58
+ posts have HTML in the body. Some sites (veteran) were WordPress-imported with
59
+ raw HTML bodies inside their `.md` files. That is legacy, not the target.
60
+ Astro renders markdown and inline-HTML identically, so a new markdown post sits
61
+ fine alongside old HTML ones. Never copy `<p>`, `<h2>`, `<table>` from an
62
+ existing post — write markdown.
63
+
64
+ Critical distinction for taxonomy fields (`category`, `tags`):
65
+ - A schema may have `category` (an enum — constrained) AND/OR `tags` (a free
66
+ array). Both are taxonomy.
67
+ - `blogWriter.approvedTags` governs **every taxonomy field**. Put approved
68
+ values in `category` AND in `tags`.
69
+ - **Do NOT copy free-form tag values from existing posts.** Many sites were
70
+ WordPress-imported with proliferated tags (every city, every appliance as a
71
+ tag). That sprawl is exactly what `approvedTags` exists to stop. Match the
72
+ field structure of old posts, ignore their messy values.
73
+ - If `category` is an enum, pick the one approved value that fits. If `tags`
74
+ is a free array, fill it from `approvedTags` only — not invented descriptors.
51
75
 
52
76
  ## Step 1 — Gather inputs (interactive)
53
77
 
@@ -79,7 +103,38 @@ options:
79
103
  - "XL (2500+ words)" — Definitive guide
80
104
  ```
81
105
 
82
- Default = the `wordCount.default` from config.
106
+ **Match the tier to the topic's natural scope — don't rubber-stamp the config
107
+ default.** A focused 2-option comparison or a single-question explainer is
108
+ usually Medium. A multi-angle definitive guide is Long or XL. Defaulting every
109
+ post to "Long" produces either thin Long posts or padded ones.
110
+
111
+ Rough mapping:
112
+ - Comparison / "X vs Y" (2-3 options) → Medium (800-1200), occasionally Long
113
+ - Single troubleshooting walkthrough → Medium (800-1200)
114
+ - How-to / multi-step guide → Long (1500-2500)
115
+ - Industry/buying guide covering many sub-topics → Long (1500-2500)
116
+ - Definitive "everything about X" → XL (2500+)
117
+
118
+ **The real quality bar is structural density + depth, NOT raw word count.**
119
+ A post passes if every H2 has 2-3 real paragraphs, there's at least one
120
+ list/table, and nothing reads thin. Never pad to hit a number — efficient,
121
+ complete writing at 1200 words beats 1800 words of filler every time.
122
+
123
+ **To make Long / XL posts genuinely long, add DEPTH, not words:**
124
+ - At least one worked example or named scenario ("a Surrey homeowner switching
125
+ from a 15-year-old gas furnace stacked CleanBC + BC Hydro to land at roughly
126
+ X after rebates") — concrete, not abstract.
127
+ - Real specifics: model families, actual process steps, real numbers inside
128
+ narratives, real local detail.
129
+ - Granular sub-sections (H3s) where a topic has genuine sub-parts.
130
+ - Show the concept applied, don't just define it.
131
+
132
+ If a Long post comes in short, the fix is "what concrete example or applied
133
+ detail is missing?" — never "add more adjectives." If after adding genuine
134
+ depth the post is still ~1200 words and complete, it was a Medium topic;
135
+ relabel it and move on. Do not pad.
136
+
137
+ `wordCount` in config is a loose sanity band, not a target to chase.
83
138
 
84
139
  ### Question 3 — Primary service + city focus
85
140
 
@@ -152,25 +207,47 @@ frontmatter shape exactly (detected in Step 0).
152
207
  (default 5-8). Each FAQ = real question a reader would search; answer is
153
208
  2-4 sentences, concise and complete.
154
209
  6. **Internal links** — `internalLinks.min` to `internalLinks.max` from config
155
- (default 3-6). Link to `/services/<slug>/`, `/service-areas/<slug>/`, and
156
- `/services/<slug>/<city>/` patterns. Don't link the same URL twice.
210
+ (default 3-6). **Target the MIDDLE of the range, not the floor.** For a
211
+ 3-6 range, aim for 4-5. Link the FIRST mention of each distinct city and
212
+ service in the body. A post that names a city six times but links it once
213
+ is under-linked — link the first occurrence. Don't link the same URL twice;
214
+ don't stuff. Patterns: `/services/<slug>/`, `/service-areas/<slug>/`,
215
+ `/services/<slug>/<city>/`.
157
216
  7. **Banned phrases** — if `bannedPhrases` is set in config, refuse them.
158
217
  Common offenders: "In today's fast-paced world…", "When it comes to…",
159
218
  "Look no further…", "In conclusion…".
160
- 8. **Pricing** — if `pricing: 'brackets-only'`, every price reference is a
161
- range with context. Never a single dollar figure. If `pricing: 'never-mention'`,
162
- omit pricing entirely.
219
+ 8. **Pricing** — applies only to `pricing: 'brackets-only'`:
220
+ - **Service-cost CLAIMS must be bracket ranges with context.** What a
221
+ job/install/repair/service costs is always a range:
222
+ "a furnace replacement typically runs $6,500 - $10,000 depending on
223
+ venting and efficiency." Never a single figure for a cost claim.
224
+ - **Narrative examples may be specific.** A hypothetical scenario, a
225
+ story about one homeowner's invoice, or a comparison walkthrough can use
226
+ a specific number because it references a *scenario*, not a price claim:
227
+ "the technician quoted $1,400 for the control board" is fine inside a
228
+ decision-making narrative. When in doubt, bracket it.
229
+ - If `pricing: 'never-mention'`, omit all pricing.
163
230
  9. **CTA** — closing paragraph includes a call-to-action. Two patterns:
164
231
  - Modal: `[text](#contact)` or trigger-class link
165
232
  - Phone: `[text](tel:+1...)` — use the actual number from `site.business.telephone`
166
233
  Best is one paragraph with both.
167
- 10. **Approved tags only.** Pick 2-3 from `blogWriter.approvedTags`. Primary
168
- tag first. **Never invent.** If nothing in `approvedTags` fits the topic,
169
- STOP and ask the user:
170
- > "None of your approved tags fit '<topic>'. Options: pick the closest fit
171
- > from <list>, OR add a new tag to `blogWriter.approvedTags` in
172
- > codejitsu.config.ts. Which?"
173
- Don't auto-add tags to the config.
234
+ 10. **Approved tags only for ALL taxonomy fields, hard cap 12.** `approvedTags`
235
+ is the single controlled vocabulary, max 12 per site. It governs both
236
+ `category` (if the schema has an enum) and `tags` (if the schema has a free
237
+ array). Put `approvedTags` values in both. Pick 2-3, primary first.
238
+ **Never invent, even if existing posts use free-form tags.** WordPress
239
+ imports often have proliferated tags (cities, appliance types) — that
240
+ sprawl is exactly what `approvedTags` prevents. Match the field *shape* of
241
+ old posts, not their values.
242
+
243
+ **If nothing in `approvedTags` fits the topic**, STOP and ask:
244
+ > "None of your approved tags fit '<topic>'. Either: (a) pick the closest
245
+ > from <list>, or (b) add a new approved tag. You currently have <N>/12.
246
+ > [If N < 12:] I can add '<proposed>' to blogWriter.approvedTags. [If N = 12:]
247
+ > You're at the 12-tag cap — to add one, tell me which existing tag to remove."
248
+
249
+ Only add a tag to the config with explicit user approval, and never exceed
250
+ 12. Don't auto-add.
174
251
  11. **Image placeholder.** Frontmatter `image` field points to where the image
175
252
  WILL live: `<imageStyle.outputDir>/<slug>.webp`. The file doesn't exist
176
253
  yet; that's fine. Run `/blog-images` to generate prompts later.
@@ -205,10 +282,26 @@ exactly. Examples:
205
282
  After writing the file, **actively run these checks** by reading the file back
206
283
  and counting. Don't skip. If any fail, fix in place before reporting Step 5.
207
284
 
208
- - Word count is within `wordCount` target (count after stripping frontmatter)
209
- - FAQ count in frontmatter is within `faqs.min`-`faqs.max`
210
- - Internal links count: grep for `](/services/` and `](/service-areas/` in body
211
- - Tags are all from `approvedTags`
285
+ - **Structural density (the real bar):** every H2 has 2-3 paragraphs; at least
286
+ one list or table; no section reads thin. This matters more than word count.
287
+ - Word count is a SOFT signal, not a gate. Do not pad to hit it. If a post
288
+ intended as Long lands short, ask "what concrete example / applied detail
289
+ is missing?" and add depth — OR accept it was a Medium topic and relabel.
290
+ Never add filler to chase a number. A dense, complete 1200-word post passes.
291
+ - FAQ count in frontmatter is within `faqs.min`-`faqs.max`.
292
+ - **Internal links — mechanical check, do this explicitly:**
293
+ 1. List every distinct service AND city you named in the body.
294
+ 2. You should have linked the FIRST mention of each (up to `internalLinks.max`).
295
+ 3. Count your actual `](/services/` + `](/service-areas/` links.
296
+ 4. If the count is below the range midpoint AND you have named-but-unlinked
297
+ services/cities, add those links now. Don't ship at the floor when you
298
+ mentioned linkable things and skipped them.
299
+ Example failure: post discusses heat pumps, furnaces, AND air conditioning
300
+ but only links two of them — link the third.
301
+ - **Every taxonomy field uses `approvedTags`**: check BOTH `category` and
302
+ `tags` in the frontmatter. If `tags` contains any value not in
303
+ `approvedTags` (e.g. a city name or appliance copied from old posts),
304
+ that's a fail — replace with approved values.
212
305
  - **No em dashes**: grep body for `—` and `–` — neither should appear
213
306
  - **No H1 in body**: first non-frontmatter line is NOT `# ...`
214
307
  - **No `bannedPhrases`**: grep body for each banned phrase
@@ -0,0 +1,231 @@
1
+ import fs from 'fs';
2
+ import path from 'path';
3
+ import { spawn } from 'child_process';
4
+ import matter from 'gray-matter';
5
+ import { loadConfig, isModuleEnabled } from '../../config/src/load.mjs';
6
+ import { c } from './format.mjs';
7
+
8
+ const PACKAGE_PLAYBOOK = '@ibalzam/codejitsu-core/modules/blog-writer/BLOG_WRITING.md';
9
+
10
+ /**
11
+ * `codejitsu blog:selftest [--topic "..."] [--model sonnet]`
12
+ *
13
+ * Spawns a COLD `claude -p` session (no memory of the kit's design), points
14
+ * it at the blog-writer playbook, has it write a throwaway post, then grades
15
+ * the output against the playbook's own rules. Reports pass/fail and deletes
16
+ * the test post. Report-only: exits 1 on failure (CI-friendly) but blocks
17
+ * nothing by default.
18
+ *
19
+ * This is the antidote to self-test bias — a fresh session reveals whether
20
+ * the playbook actually drives behaviour, not whether the author follows
21
+ * rules they already know.
22
+ */
23
+ export async function runBlogSelftest({ topic, model = 'sonnet' } = {}) {
24
+ const cwd = process.cwd();
25
+
26
+ let config;
27
+ try { config = await loadConfig(cwd); }
28
+ catch (err) {
29
+ console.error(c.red(`✗ ${err.message}`));
30
+ process.exit(1);
31
+ }
32
+
33
+ if (!isModuleEnabled(config, 'blogWriter')) {
34
+ console.error(c.red('✗ blogWriter not configured in codejitsu.config.'));
35
+ console.error(' Run `codejitsu blog:init` and add the blogWriter block first.');
36
+ process.exit(1);
37
+ }
38
+ const bw = config.blogWriter;
39
+
40
+ if (!(await commandExists('claude'))) {
41
+ console.error(c.red('✗ `claude` CLI not in PATH. Install Claude Code to run selftest.'));
42
+ process.exit(1);
43
+ }
44
+
45
+ // Default topic: weave service[0] + location[0] so internal links are possible.
46
+ const svc = bw.services?.[0] ?? 'your main service';
47
+ const loc = bw.locations?.[0] ?? 'your main city';
48
+ const testTopic = topic ?? `${svc} considerations for ${loc} homeowners this season`;
49
+
50
+ const slug = `__selftest-${Date.now()}`;
51
+ const contentDir = path.resolve(cwd, (typeof config.blog === 'object' && config.blog.contentDir) || 'src/content/blog');
52
+ const postPath = path.join(contentDir, `${slug}.md`);
53
+
54
+ console.log(c.bold(`\nCodejitsu blog:selftest\n`));
55
+ console.log(`Model: ${model}`);
56
+ console.log(`Topic: ${testTopic}`);
57
+ console.log(`Output: ${path.relative(cwd, postPath)} (deleted after grading)`);
58
+ console.log(c.gray('\nSpawning a cold claude -p session… (~$0.80 on sonnet, 2-6 minutes)\n'));
59
+
60
+ const prompt = buildPrompt({ testTopic, slug });
61
+
62
+ let envelope;
63
+ try {
64
+ const out = await runCmd('claude', [
65
+ '-p', prompt,
66
+ '--allowedTools', 'Read Glob Grep Write Edit Bash',
67
+ '--model', model,
68
+ '--output-format', 'json',
69
+ ], 600_000);
70
+ if (out.code !== 0) {
71
+ console.error(c.red(`✗ claude exited ${out.code}`));
72
+ console.error(out.stderr.slice(0, 400));
73
+ cleanup(postPath);
74
+ process.exit(1);
75
+ }
76
+ envelope = JSON.parse(out.stdout);
77
+ } catch (err) {
78
+ console.error(c.red(`✗ Could not run claude: ${err.message}`));
79
+ cleanup(postPath);
80
+ process.exit(1);
81
+ }
82
+
83
+ if (envelope.total_cost_usd) {
84
+ console.log(c.gray(`Cost: ~$${Number(envelope.total_cost_usd).toFixed(3)}\n`));
85
+ }
86
+
87
+ if (!fs.existsSync(postPath)) {
88
+ console.error(c.red(`✗ Cold Claude did not write the expected file: ${path.relative(cwd, postPath)}`));
89
+ console.error(c.gray('Its response was:'));
90
+ console.error(c.gray((envelope.result ?? '').slice(0, 500)));
91
+ process.exit(1);
92
+ }
93
+
94
+ const results = gradePost(postPath, bw);
95
+ printGrade(results);
96
+ cleanup(postPath);
97
+
98
+ const fails = results.filter((r) => r.status === 'fail').length;
99
+ console.log('');
100
+ console.log(fails === 0
101
+ ? c.green('Selftest passed — the playbook drove a clean post from a cold session.')
102
+ : c.red(`Selftest found ${fails} failure(s) — tighten the playbook and re-run.`));
103
+ process.exit(fails > 0 ? 1 : 0);
104
+ }
105
+
106
+ function buildPrompt({ testTopic, slug }) {
107
+ return `You are running the /blog command for this Codejitsu site. Open ${PACKAGE_PLAYBOOK} (in node_modules) and follow it top-down to write ONE blog post.
108
+
109
+ This is a NON-INTERACTIVE selftest run. SKIP the AskUserQuestion step and use:
110
+ - Topic: ${testTopic}
111
+ - Pick the post type + length tier that genuinely fit the topic
112
+ - Use the next sensible future publish date
113
+ - IMPORTANT: write the file to exactly this slug so it can be graded + cleaned up:
114
+ src/content/blog/${slug}.md
115
+
116
+ Read codejitsu.config.ts for tone, services, locations, approvedTags, and all rules. Read src/content.config.ts for the taxonomy fields. Read ONE existing post to match the frontmatter SHAPE (but write the body in clean markdown, and use only approvedTags for category + tags). Run the playbook's Step 4 verify checks and fix any failures before finishing.`;
117
+ }
118
+
119
+ function gradePost(postPath, bw) {
120
+ const raw = fs.readFileSync(postPath, 'utf8');
121
+ let data, body;
122
+ try {
123
+ const parsed = matter(raw);
124
+ data = parsed.data;
125
+ body = parsed.content;
126
+ } catch (err) {
127
+ return [{ status: 'fail', label: 'Frontmatter parses', detail: err.message }];
128
+ }
129
+
130
+ const results = [];
131
+ const add = (ok, label, detail) =>
132
+ results.push({ status: ok ? 'pass' : 'fail', label, detail: ok ? undefined : detail });
133
+ const addWarn = (ok, label, detail) =>
134
+ results.push({ status: ok ? 'pass' : 'warn', label, detail: ok ? undefined : detail });
135
+
136
+ // Em dashes
137
+ const emDashes = (body.match(/[—–]/g) ?? []).length;
138
+ add(emDashes === 0, 'No em dashes', `found ${emDashes}`);
139
+
140
+ // No H1 in body
141
+ const firstLine = body.split('\n').find((l) => l.trim().length > 0) ?? '';
142
+ add(!/^#\s/.test(firstLine.trim()), 'No H1 in body', `body starts with: ${firstLine.slice(0, 50)}`);
143
+
144
+ // Markdown body, not HTML
145
+ add(!/^\s*<(p|h[1-6]|div|ul|ol|table)[\s>]/i.test(body.trimStart()),
146
+ 'Body is markdown (not HTML)', 'body starts with an HTML block tag');
147
+
148
+ // FAQ count
149
+ const faqMin = bw.faqs?.min ?? 5;
150
+ const faqMax = bw.faqs?.max ?? 8;
151
+ const faqCount = Array.isArray(data.faqs) ? data.faqs.length : 0;
152
+ add(faqCount >= faqMin && faqCount <= faqMax,
153
+ `FAQ count in ${faqMin}-${faqMax}`, `got ${faqCount}`);
154
+
155
+ // Tags governance
156
+ const approved = new Set(bw.approvedTags ?? []);
157
+ if (approved.size > 0) {
158
+ const used = [];
159
+ if (typeof data.category === 'string') used.push(data.category);
160
+ if (Array.isArray(data.tags)) used.push(...data.tags);
161
+ const bad = [...new Set(used.filter((t) => !approved.has(t)))];
162
+ add(bad.length === 0, 'All tags/category in approvedTags',
163
+ `not approved: ${bad.join(', ')}`);
164
+ } else {
165
+ results.push({ status: 'warn', label: 'approvedTags configured', detail: 'none set' });
166
+ }
167
+
168
+ // Internal links
169
+ const linkMin = bw.internalLinks?.min ?? 3;
170
+ const links = (body.match(/\]\(\/(services|service-areas)\/[^)]+\)/g) ?? []).length;
171
+ addWarn(links >= linkMin, `Internal links ≥ ${linkMin}`, `got ${links}`);
172
+
173
+ // Banned phrases
174
+ const banned = bw.bannedPhrases ?? ["In today's fast-paced world", 'When it comes to', 'Look no further', 'In conclusion'];
175
+ const hit = banned.filter((p) => body.toLowerCase().includes(p.toLowerCase()));
176
+ add(hit.length === 0, 'No banned phrases', `found: ${hit.join('; ')}`);
177
+
178
+ // At least one list or table
179
+ const hasList = /^[-*]\s/m.test(body) || /^\d+\.\s/m.test(body) || /^\|.+\|/m.test(body);
180
+ add(hasList, 'At least one list or table', 'none found');
181
+
182
+ // Word count sane (not absurdly thin)
183
+ const words = body.split(/\s+/).filter(Boolean).length;
184
+ addWarn(words >= 700, 'Body ≥ 700 words (not thin)', `got ${words}`);
185
+ results.push({ status: 'info', label: `Body word count: ${words}` });
186
+
187
+ // Pricing brackets (only if brackets-only)
188
+ if (bw.pricing === 'brackets-only') {
189
+ // Heuristic: flag standalone $N that is NOT part of a "$N - $N" range or a narrative.
190
+ // We only hard-check the count; narrative examples are allowed, so this is a warn.
191
+ const standalone = (body.match(/\$[0-9][0-9,]*(?!\s*-\s*\$)/g) ?? []).length;
192
+ results.push({ status: 'info', label: `Standalone $ figures: ${standalone} (narrative examples allowed; service-cost claims should be ranges)` });
193
+ }
194
+
195
+ return results;
196
+ }
197
+
198
+ function printGrade(results) {
199
+ console.log(c.bold('Grade:'));
200
+ for (const r of results) {
201
+ const icon = r.status === 'pass' ? c.green('✓')
202
+ : r.status === 'warn' ? c.yellow('!')
203
+ : r.status === 'info' ? c.gray('i')
204
+ : c.red('✗');
205
+ console.log(` ${icon} ${r.label}${r.detail ? c.gray(' — ' + r.detail) : ''}`);
206
+ }
207
+ }
208
+
209
+ function cleanup(postPath) {
210
+ try { if (fs.existsSync(postPath)) fs.unlinkSync(postPath); } catch {}
211
+ }
212
+
213
+ function commandExists(cmd) {
214
+ return new Promise((resolve) => {
215
+ const p = spawn('which', [cmd], { stdio: 'ignore' });
216
+ p.on('close', (code) => resolve(code === 0));
217
+ p.on('error', () => resolve(false));
218
+ });
219
+ }
220
+
221
+ function runCmd(cmd, args, timeoutMs) {
222
+ return new Promise((resolve, reject) => {
223
+ const proc = spawn(cmd, args, { stdio: ['ignore', 'pipe', 'pipe'] });
224
+ let stdout = '', stderr = '';
225
+ proc.stdout.on('data', (d) => { stdout += d; });
226
+ proc.stderr.on('data', (d) => { stderr += d; });
227
+ const t = setTimeout(() => { proc.kill(); reject(new Error('timed out')); }, timeoutMs);
228
+ proc.on('error', (e) => { clearTimeout(t); reject(e); });
229
+ proc.on('close', (code) => { clearTimeout(t); resolve({ code, stdout, stderr }); });
230
+ });
231
+ }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@ibalzam/codejitsu-core",
3
- "version": "0.11.1",
3
+ "version": "0.12.0",
4
4
  "type": "module",
5
5
  "description": "Shared core for Codejitsu Astro sites — reusable code and Claude-facing instructions for blog, SEO, images, deploy, and llms.txt.",
6
6
  "keywords": [