docs(pipeline): teach LLM to avoid 7 patterns audit found across last 10 posts

[pftg]

Summary

After driving PR #322's AI-detector score 52 → est. 22-28 across 3 cold-pass sweeps, an empirical audit on the 10 most recent posts surfaced patterns the new pipeline (PR #319) didn't catch. This PR adds those patterns to the voice guide and writer prompt as teaching (concrete BAD/GOOD pairs, surface-feature descriptions). No tool recommendations, no programmatic gates - the LLM learns from the patterns.

What the audit found (last 10 posts, ~22k words)

Pattern	Posts affected	Pipeline catches?
Long paragraphs (>3 sentences OR >70w)	5/10 (TDD/XP cluster)	❌
Anaphora pairs (same first-word sentences in paragraph)	10/10	❌
"The..." opener density (>20%)	6/10	❌
"We..." opener stamping	TDD/XP cluster	❌
Mannerism-noun saturation (9-18 per post, cap 3)	6/10	partial
Architecture enumeration in body	TDD/XP cluster	❌
Formulaic founder-anecdote hook (Priya/Marcus/Sarah opener)	5/5 founder-pain posts	❌
Definitional cadence	1-2/10	✅
Em dashes	0/10	✅
"Not X but Y"	0-1/10	✅

The gap: pipeline catches lexical/sentence-level tells; misses paragraph-level and post-level structure tells.

What changed

docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md

Added 6 new rows to the "Banned structural patterns" table:

• Anaphora pairs
• "The..." opener density (cap 20%)
• "We..." opener density (cap 15%)
• Long paragraphs (3 sentences AND 70 words working cap)
• Architecture enumeration in body
• Formulaic founder-anecdote hook

Extended the existing Mannerism-noun refrain row:

• Cap raised from 3 to 5 (acknowledges some posts genuinely need labels)
• Watch list extended to include cadence, rhythm, contract, gate, gatekeeper, layer, round, review, pass

docs/workflows/blog-pipeline.md

Added to STEP 4b (cadence quotas):

• Paragraph cap = 3 sentences AND 70 words
• Anaphora rule
• "The..." and "We..." opener-density caps
• Mannerism cap raised to 5, watch list extended
• Architecture-enumeration ban
• Hook variety guidance

Added to STEP 4c (BAD/GOOD pairs):

• Anaphora pair rewrite
• Architecture enumeration rewrite
• Long-paragraph split example
• Mannerism-noun saturation rewrite
• Formulaic founder-anecdote hook with alternative opener shape
• "We" opener stamping with rewrite

Approach: TEACHING, not gating

Per request: no tool recommendations (no Pangram, GPTZero, grep scripts), no programmatic gates. The LLM reads the patterns + BAD/GOOD pairs and learns what good writing looks like. The cold-eyes Agent J (already in pipeline) catches what the writer misses.

Test plan

• bin/hugo-build — passes
• Next blog post written under updated pipeline measures whether the new patterns are caught at draft time
• If next post still scores >35 on first read, the audit's recommendation about programmatic gates becomes more compelling

🤖 Generated with Claude Code

[coderabbitai]

Warning

Rate limit exceeded

@pftg has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 15 minutes and 17 seconds before requesting another review.

To keep reviews running without waiting, you can enable usage-based add-on for your organization. This allows additional reviews beyond the hourly cap. Account admins can enable it under billing.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 33dec257-06d3-4323-b9cd-775171648954

📥 Commits

Reviewing files that changed from the base of the PR and between f1aeb21 and f6a1c5f.

📒 Files selected for processing (2)

• docs/90-99-content-strategy/strategy-analysis/90.11-voice-guide.md
• docs/workflows/blog-pipeline.md

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch chore/voice-guide-anaphora-paragraph-rules

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Review rate limit: 0/1 reviews remaining, refill in 15 minutes and 17 seconds.}

_{Comment @coderabbitai help to get the list of available commands and usage tips.}