b05ffbfbd6d0dfce44fbe702eca276603e46485c
7
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  dennisthiessenandClaude Opus 5
|
f2a1229701 |
fix: clarify the Bosch 24/7 constraint; tidy build artefacts
BS-1 operating constraint - the F07 open item, resolved by the user The user clarified what the 24/7 constraint actually is: a defect in the software could, worst case, halt production - an outcome that must never occur. That is a FAILURE-CONSEQUENCE constraint, and the submitted NVIDIA cover letter states something different: "a production line that could not be stopped for a deployment" describes a scheduling window. The distinction matters in both directions - the letter is inaccurate, and the truth is the stronger claim. - BS-1 scope now records the real constraint and how to write it. - Both variants of the letter's phrasing are added to BS-1 `forbidden`, so the scanner catches them. Verified: it now flags the submitted letter. - BS-1 evidence filled from a real source (this clarification) rather than left unrecorded, including ownership - integration and operation, not model work. - The submitted letter is NOT rewritten (F17). experience_bosch.md records the resolution and the wording to use instead. Scanner false positive fixed: the employer-date check now runs on CVs only. A cover letter naming an employer in prose is not claiming a date. Cleanup, authorised by the user - 136 LaTeX build artefacts removed (already gitignored, regenerable). - 7 superseded intermediate QA renders removed, 6 of them in NATO_AI_ENGINEER which carried nine PDFs for two documents. - Nothing evidential touched: 36 submitted PDFs, 41 e2e deliverables, 44 sources, 54 session/critique files and 20 JD files are intact. The Kraken PDF was checked before deleting and turned out to be the saved job description, not a render. - output/ 16M -> 13M. 37/37 tests pass. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qej5SBvhcbGXHwLqNWiskM |
||
|
dennisthiessenandClaude Opus 5
|
eecdbdade6 |
fix(F17): a PDF in a folder is not proof the package can be rebuilt
Build sweep run this session: 48 sources, 46 built, 2 failed. Infineon_AI_Engineer's cover letter fails on font expansion (the font is not scalable); cv_template.tex fails on unfilled placeholders, which is expected for the academic legacy path. Overfull boxes: Microsoft ISE 25.33pt, Isovalent 0.88pt. All historical, all left alone. New: resume_builder/helpers/package_manifest.py records, per application package, what a rebuild would ACTUALLY do - sources, class-file hashes, PDFs present, and the reasons a rebuild would not reproduce the submitted document. It is descriptive; nothing in it authorises rewriting a submitted file. What it found across 21 packages: - 17 are DATE-DEPENDENT. The cover letters use oday, so recompiling a historical package silently restamps it with today's date. - 1 has an INCOMPLETE SOURCE DISTRIBUTION: output/NATO_AI_ENGINEER ships no resume.cls. It builds on this machine only because a copy exists elsewhere, and the shipped PDF hid that. Not repaired by copying the newest class in - the repo carries THREE different resume.cls variants (13/7/1 copies) and older macros differ, so that "fix" could silently change a historical document. - 1 has no PDF at all, so nothing is reviewable without a build. Classification: 7 application packages were in no reuse category. Added to historical_revalidate, and the register now records the default explicitly - anything unlisted is treated as revalidate-before-reuse. Being recent is not the same as being verified; none of the seven was re-read for this. Prevention rather than repair: oday removed from the cover-letter template so new packages stop adding to the pile. A test now fails if it comes back, if the manifest goes stale, or if a package is left unclassified. Also: scan_document gained a template mode. A placeholder in a TEMPLATE is what makes it a template; the defect is one surviving into a generated document. 37/37 tests pass; counter-proof still catches all 9. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qej5SBvhcbGXHwLqNWiskM |
||
|
dennisthiessenandClaude Opus 5
|
3519b2ffe7 |
fix(F08-F10): machine-checkable evidence, honest templates, demoted GPA
F08 - evidence metadata - Only 1 of 26 claims carried structured source/last_verified/ownership, which is why recency and reference checks could not run. Added an evidence_schema block and an evidence block on all 26 claims. - 8 dates were recovered from each claim's OWN recorded text; the other 18 are written as the literal "unrecorded". Nothing was inferred. "unrecorded" means nobody wrote it down, NOT that the claim is unevidenced - most carry evidence in their free-text scope. - Coverage is reported as an advisory warning, never a gate: failing on it would push a future session to invent dates to clear the run. F09 - templates - resume_template.tex asserted "12+ years of production software experience". It does not hold: canonical employment starts 2014-11 (~11y10m) and the VISIBLE chronology starts 2015-05 (~11y4m) because Capgemini is omitted. Now "over ten years", with the arithmetic and the no-date-math rule inline. - cv_template.tex still carries page targets, rendered-line budgets and 2L/3L variants the project dropped. Banner-marked as the ACADEMIC legacy path so it is not reached for whenever a posting says "CV". F10 - US GPA - The modified Bavarian formula converts FOREIGN grades INTO the German system. Inverting it to produce a US GPA is not a recognised method, and the TUM source calls even its forward calculator orientation only. - Fields renamed to final_grade_us_4pt_orientation_only with that caveat. The German grades remain the canonical facts. If a recipient publishes its own conversion, use theirs and record which was used. Validator PASS (1 advisory), 31/31 tests. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qej5SBvhcbGXHwLqNWiskM |
||
|
dennisthiessenandClaude Opus 5
|
fc5f3e49ba |
fix: review findings F03-F06 — quotas, rule hierarchy and a validator that checks
F03 — application records
- Quota already removed with F02's tracker; verified uncapped and that
"interview reached" survives a later withdrawal.
- Aker BP: application date confirmed 2026-07-29 by the user, settling the
one-day conflict. Its Adjacent-despite-79 classification is kept, NOT
reclassified: the user does not remember the rationale and judged it
immaterial on a closed role, and silently reclassifying to make the
arithmetic tidy is the exact objection the review raised. Recorded as an
explicit override reason with a rubric version instead.
F04 — rule hierarchy
- Three levels stated in AGENTS.md and CLAUDE.md: immutable evidence rules,
standing user preferences, market/employer presentation.
- Deliverables separated from QA. The .tex stays the artefact of record; you
still compile, into the application's own output/ folder, because the user
cannot approve what he cannot open.
- `resume_reference.md` said "Capgemini/Bundeswehr: include only when directly
relevant", contradicting the standing never-include rule. Split and fixed.
F05 — a green validator now means something
- Per-claim `forbidden` wording is enforced. 14 of 26 claims carried such a
list and NONE of it was checked, which is why "owned the full ML lifecycle"
passed. Matching tolerates inserted adjectives: the real regression read
"Built a decentralized Data Mesh" and one word defeated the literal list.
- Added identity (e-mail/phone), employer-date, placeholder and
document-exclusion checks. Capgemini's omission rule now has a canonical
home in claims.json instead of living only in memory and config.md.
- Registered Groovy and Snowflake as forbidden skills. Both were absent from
claims.json entirely, so "forbidden" existed in prose with nothing enforcing
it.
- PASS no longer reads as a truth certificate: it prints what it did not check
— scope implications, bullet-to-claim support, whether a number is real.
F06 — the scanner stopped fighting the documents
- LaTeX comments are no longer claims; `% never claim Terraform` was a finding
before, which taught the workflow to write around the checker.
- Markup, accent macros and umlauts are normalized, so `\textbf{Snow}flake`
cannot smuggle a forbidden skill past a substring test and
`Universit\"at der Bundeswehr M\"unchen` is recognisably a school.
- Honest denials pass in English and German.
- The triple-list cadence hint knew only English "and", so it silently did
nothing on every German package. Advisory only, and now labelled as such.
Verification
- The six counter-proofs from the report are caught: 9 errors on one synthetic
document, where the old scanner found none.
- All six packages built under current rules scan clean — SBB, NVIDIA, RUAG,
Kdo Cy, Citadel, Aker BP. Four false positives of my own were found and fixed
on the way there; each is now a regression test.
- 13 older packages do report findings. Those look like genuine historical
violations and are deliberately NOT fixed here: F17 governs archives and they
must not be silently rewritten.
- `tests/test_document_scanner.py` added. Full suite 31/31 pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qej5SBvhcbGXHwLqNWiskM
|
||
|
dennisthiessenandClaude Opus 5
|
afdef7eb94 |
refactor: shared agent workflows (F01) and one application registry (F02)
Checkpoint for the 2026-09-11 project review remediation. Work by a previous agent that ran out of context; committed unchanged as a rollback point before the remaining findings land on top of it. F01 — the two agent directories had drifted into different processes - Six shared workflows under `resume_builder/reference/workflows/`; both `.agents/skills/` and `.claude/skills/` are now thin entrypoints that route to the same file instead of carrying full copies. - The `.claude` copies were the stale side: fixed bullet line variants and character budgets that the newer process had already dropped. Removing them is the point of the change, not collateral. - Parity/link regression test added; corrected the top-level file map, which pointed at a directory that does not exist. F02 — application status was stored in four places and disagreed - New registry `job_scout/state/applications.json` (26 records) with `application_records.py` (schema, validation, transactional write) and `application_tracker.py` (add/update, no quota). - AGENTS.md and CLAUDE.md no longer keep their own status tables; both point at `current_applications.md`, generated from the registry. - Interview-reached is now independent of outcome, so a withdrawal after an interview can record both facts. The old tables are preserved verbatim in `job_scout/state/history/application_status_before_2026-09-11.json`. Verification: `python -m unittest discover -s tests -v` — 18/18 pass. Spot-checked that the detailed SBB and Kdo Cy notes survived the migration into the registry rather than being flattened. Known unresolved, deliberately not invented (see FIX_PROGRESS.md): Aker BP's application date and its Adjacent-despite-79 classification. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Qej5SBvhcbGXHwLqNWiskM |
||
|
dennisthiessenandClaude Opus 5
|
92a81a7dcb |
feat(resume): check bullet cadence variety, warn-only
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
|
||
|
dennisthiessen
|
fe5f24704f | feat: rebuild evidence-first application workflow |