dennisthiessenandClaude Opus 5
92a81a7dcb
An audit of all 18 packages in output/ (368 bullets) found one rhythm
running through every document: 45% of bullets used the same "X, Y and
Z" triple, and the SBB package reached 85% - 11 of 13 bullets, every
Swisscom and Bosch line - plus two adjacent bullets both opening
"Build and...".
This is a style finding, not a truth finding. Every bullet was accurate.
The corpus is clean on the axes that actually signal generated text: no
AI vocabulary (0 hits for leverage/spearheaded/robust/passionate and 56
others across 35 documents), prose em-dashes at 0.08/bullet, and PDF
metadata carrying nothing but MiKTeX pdfTeX with empty Author/Title
(0 AI tokens and 0 generator-term leaks across 69 PDFs). What is left is
cadence: ten bullets sharing one three-beat rhythm read as machine-made
even when nothing in them is false.
Guarded deliberately so it cannot do harm. cadence_checks() emits WARN
and never ERROR, the critique deduction caps at 1 point, and both the
reference and the docstring state that no claim, scope or hedged verb
may be bent to satisfy rhythm. An anti-monotony rule with teeth would be
worse than the problem - it would pressure a future run into loosening a
scoped claim to vary a sentence.
Thresholds are calibrated on the corpus, not guessed. Position length
spreads are bimodal (1-9 words, then 15-17), so the check flags a spread
of <=4, the tight tail at ~28% of positions; the first draft used <=6
and flagged the median. Also fixed the opening-verb extractor, which
read \textbf{Owned ...} as the word "textbf" and produced six false
positives on one document. Per-package warnings now run 0-3.
Docs: resume_reference.md 7a + verification step, critical_rules.md 7a,
critique_framework.md mechanics row, CLAUDE.md corrections log.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MHtzyTKBcg6BWhD5qFegtK
4.3 KiB
Application Critique Framework
Never collapse candidate-role fit and document polish into one score.
1. Hard-Gate Audit
Classify each JD requirement as Direct, Adjacent, Gap, or Constraint.
Flag a NO-GO when:
- A minimum/required qualification is a Gap.
- A title-defining capability is not Direct.
- The level requires absent scope, such as people management, strategic external-account ownership, or company-wide architecture authority.
- Location, authorization, clearance, travel, or mandatory relocation is unresolved.
An employer's general "apply anyway" message does not erase a title-defining or minimum gap.
2. Evidence Fit Score — 0 to 100
Score the candidate-role pairing before reading tailored prose.
| Dimension | Weight | Question |
|---|---|---|
| Required qualifications | 35 | How many are Direct? |
| Core responsibilities | 25 | Has Dennis performed the actual function? |
| Level and ownership | 15 | Do title, decisions and leadership match? |
| Recency and depth | 10 | Is the strongest evidence current and substantial? |
| Domain/tool transfer | 10 | Are missing tools genuinely substitutable? |
| Practical constraints | 5 | Are location, language, authorization and travel workable? |
- 75--100 Core: apply when no hard gate fails.
- 60--74 Adjacent: selective; normally no more than one serious preferred gap.
- Below 60 Stretch: normally do not cold-apply.
- Any hard-gate failure overrides the number.
3. Document Quality Score — 0 to 100
| Dimension | Weight | Question |
|---|---|---|
| Truth and provenance | 25 | Does every claim match canonical scope? |
| Information hierarchy | 20 | Are employer, title, dates and strongest evidence obvious? |
| Bullet evidence and impact | 20 | Are bullets specific and scoped without invention? |
| Relevance and terminology | 15 | Does it cover the JD naturally? |
| Skills evidence | 10 | Is every skill evidence-backed and interview-ready? |
| Mechanics/readability | 10 | Does it compile, parse and read cleanly? Includes cadence variety — read the bullets in sequence and count "X, Y and Z" triples, repeated opening verbs and length clustering per position (resume_reference.md §7a). Deduct at most 1 pt; it is a style flaw, never a truth flaw. |
Truth/provenance below 8/10 is an automatic document failure.
4. Channel Strength — Separate Rating
| Rating | Evidence |
|---|---|
| Strong | Referral/internal sponsor plus targeted hiring-team contact |
| Moderate | Recruiter conversation, known hiring manager or warm introduction |
| Weak | Tailored cold application only |
Do not mix channel strength into either score. A polished cold application remains a weak channel.
5. Competitive Read
Describe the obvious-fit competitor and compare direct function, cloud/tool ecosystem, domain, level/scope, and access to the team. State Dennis's real advantage and the competitor's real advantage. Do not value a bridge as highly as direct evidence.
6. Reader Sequence
Use verdicts, not invented probability percentages:
- Parser/ATS: extraction and relevant terms.
- Recruiter glance: Forward / Borderline / Reject, with one reason.
- Hiring manager: Interview / Maybe / No, focused on functional evidence.
- Technical reviewer: credibility and likely first challenge.
7. Claim Audit
Audit every material summary, skills and high-value bullet claim:
| Claim | Canonical ID/evidence | Direct/Adjacent | Safe? | Fix |
|---|
Run validate_resume_system.py --document first. Any validator failure is Tier 1.
8. Cover Letter
If present, assess whether the letter was useful under the decision gate, adds information, uses current first-party hooks, and contains only canonical/resume claims. Consider whether deleting it would improve the application.
If absent, assess whether the resume stands alone. Do not penalize an optional missing letter.
9. Required Critique Output
- Fit verdict and hard-gate failures.
- Evidence Fit score with dimension table.
- Document Quality score with dimension table.
- Channel Strength rating.
- Competitive read and reader sequence.
- Claim audit.
- Tier 1 truth/fit fixes and Tier 2 presentation improvements.
- Cover-letter decision/critique.
- Mechanical verification.
Never issue a single overall score or call a package submit-ready solely because Document Quality is high.