Resume workflow: operator reference (beta)
People use Create a resume. The assistant follows the resume setup contract when direction or preferences need clarification. Its saved brief feeds selection and planning below.
Install Career Evidence Core first. Both components use the same workspace and versioned career schema; no import or conversion is needed when adding the app. The factual record, supported interpretations and preferences belong to the core.
For a document, use make-resume (or make-interview-brief for private preparation).
See Editorial memory for briefs, ranked selections, scoped
feedback and exact provenance. Generation uses the selected safe evidence view;
private interview preparation deliberately reads the full private pack.
The application owns data/roles, data/briefs, data/selections, data/plans, and
reviews/decisions. Its generated Markdown, exports and review sidecars belong
under outputs. Deleting outputs must not erase selection choices or feedback.
Application records reference stable core IDs and exact pack hashes. Schema 1.3
and 1.4 packs remain supported. Legacy editorial.py status, migrate and
bind-strength commands forward to career_core.py.
A wording change is an output edit. A changed inclusion preference is a scoped editorial decision. A new achievement, ownership correction or measurement is core evidence work: record the source answer, write a new pack version and reassess affected interpretations. An application cannot upgrade an evidence status or silently turn its own prose into a source.
Before delivering, run representation and integrity review, then a cold recruiter screen. Review exact final bytes; regenerate manifests after edits. Every resume requires PDF, TXT and DOCX exports. Check final PDF layout and verify recovered text across all three formats. Report shortcomings independently: a faithful resume may still be weak, and a well-written one may target an unsupported role.
The add-on includes Makefile.resume for its deterministic checks. The developer
checkout has make check-resume and make test-releases as well. Model scenarios
are separate, consume tokens and require the locally configured Claude CLI.
See Resume authoring for the full planning and review flow and Exports for prerequisites and failure handling.
JSON Resume projection
The private career pack can be projected into the separate JSON Resume presentation format:
python3 scripts/export_resume_json.py --audience public -o outputs/resume.json
This lossy export omits ineligible material, retains separate position rows for promotions, converts recorded dates and records its canonical source. It does not replace evidence, sources or review history in the career pack. Third-party themes need their own layout checks. Use Core's private JSON export for a lossless copy of the accepted pack.