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
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
dennisthiessenandClaude Opus 5