Tailored CV per job¶
Optional. With a curated bullet pool in place, prepare_application also produces a one-page PDF of your CV tailored to the posting, built from bullets you wrote and rendered through your own LaTeX template.
How the tailored CV is built¶
When a curated bullet pool exists — cv-pool.yaml in MOONLIGHTER_HOME, or wherever cv.pool in config.yaml points — prepare_application also tailors your CV to the posting: one LLM call selects and orders bullets from your pool and writes a short summary and a technical-expertise line from your profile (it can also answer "the base CV already fits" and change nothing), and the result renders through your own LaTeX template and compiles with pdflatex when installed.
- Always one page. The prompt carries the budget, and the orchestrator drops the least relevant bullets until pdflatex reports a single page.
- Always plain Latin text. A model field carrying an emoji, symbol or non-Latin script is replaced whole by your curated text, never edited.
- Grouped roles. Consecutive roles at the same company render as one grouped block with the overall date span, instead of two separate entries.
- Your bullets, not the model's. Every bullet comes from your pool; the model chooses and orders them (translating them for a Portuguese posting), and writes the summary and technical-expertise line under instructions to use only your profile and never inflate a claim. The sheet always tells you to review the generated PDF before uploading.
Without a pool file no CV is generated and no extra LLM call is made — but prepare_application does offer to draft you one, returning that offer instead of a sheet, until you either bootstrap a pool or skip it (see Starting a pool).
Configuration and cache¶
Config keys: cv.pool, cv.template_dir (holding cv-template.en.tex, optionally cv-template.pt.tex for Portuguese postings), cv.generated_dir (default cv-generated under MOONLIGHTER_HOME). Defaults for each are in Configuration.
Each job's result is cached under <generated_dir>/<job_id>/, so no job is ever generated twice — after editing your pool or template, delete that directory to have the next prepare_application regenerate that job's CV.
Starting a pool¶
No pool file yet? prepare_application offers to draft one from your profile.yaml whenever it's missing (or run moonlighter-apply bootstrap-cv from a shell) — a first draft you review and edit. It is built from the generic example pool and example template, which double as the schema reference if you'd rather hand-write a pool instead. Both ship inside the installed package too, under moonlighter/application/cvgen/templates/, so a checkout is not required to read them.
In Claude, bootstrap_cv_pool drafts the pool and skip_cv_bootstrap declines the offer for good (see MCP tools).