/resume

description: Generate a tailored resume and cover letter from a job description, score both, create DOCX files, and update the tracker.

Resume Builder — Native Four-Role Team + Triple Scoring

Generate a tailored resume AND cover letter using Codex parallel tool execution with scoring-aware optimization. Every editorial decision maps to ATS (7 components) and HR (6 factors) scoring weights.

Job Description

$ARGUMENTS

CANDIDATE-FIT PREFLIGHT (MANDATORY FIRST GATE)

Before scorer startup, research, resume development, role/team invocation, output or application-directory creation, DOCX generation, or tracker work, resolve `master_resume_path` from `config.json` and save this exact job description to a private temporary UTF-8 file. Screen only that configured master resume—never a previously tailored resume. Generate one safe `run_id`, one safe `case_id`, and one strict ISO `as_of_date`, then run:

`python candidate_fit_preflight.py --resume <configured-master-resume> --job-description <private-exact-JD.txt> --run-id <run_id> --case-id <case_id> --as-of-date <YYYY-MM-DD> --json`

Require exit `0` plus a valid `candidate-fit-policy-v3` report bound to the same IDs, date, master SHA-256, and exact-JD SHA-256. Canonically hash it as `candidate_fit_report_digest`. Proceed only when `threshold` is exactly `70.0`, `score >= 70`, `extraction_trustworthy` is true, `hard_knockouts` is empty, `passed` is true, and `codes` is empty. Exit `1`, a score below 70 (including 60–69), or any hard knockout is `REJECTED:CANDIDATE_FIT`; stop with no team/role invocation and no application directory, draft, DOCX, or tracker mutation. Exit `2` or an unavailable, malformed, stale, non-canonical, or digest-mismatched report is `FAILED:CANDIDATE_FIT_PREFLIGHT` and fails closed. There is no automatic or manual workflow bypass. ATS/HR baselines remain advisory and cannot override candidate fit.

Instructions

You are the coordinator, not the resume author. The user has provided a job description above. You will: 1. Pass the fixed candidate-fit gate against the configured master resume 2. Internalize both advisory scoring engines before writing a single word 3. Deconstruct the JD into a scoring blueprint 4. Delegate drafting to the native `resume-writer` with least-authority context 5. Diagnose gaps by component weight, not guesswork 6. Require an independent `resume-auditor`, bounded `resume-editor` corrections, and three deterministic authorization votes

---

GLOBAL CONSTRAINTS (read first, enforce always)

**Editorial priority (never invert):** Authenticity → Human voice → HR impact → evidence coverage.

• NEVER change job titles, company names, dates, education, publications, certifications, or memberships
• NEVER add parenthetical qualifiers to job titles — titles must match the master resume exactly, with no additions or removals.
• NEVER use `**bold**` markdown in `.md` files — the DOCX generator handles bold automatically
• NEVER exceed 2 appearances of any single keyword across the entire resume
• NEVER keyword-stuff experience bullets — Core Competencies is the primary ATS keyword home; bullets stay factual and human
• NEVER open bullets with AI-cliché verbs (spearheaded, leveraged, orchestrated, championed, …) — use plain strong verbs (Led, Built, Wrote, Cut, Reviewed)
• Publications & Education: Keep EXACTLY as in master resume — zero modifications
• DOCX and tracker finalization must use only the receipt-validating authorized wrappers specified in Phase 5; never call lower-level generators or tracker mutations directly
• Acceptance targets: **Evidence Match ≥ 75%**, **zero must-haves with NO_EVIDENCE_FOUND**, eligibility not FAIL, **Human Voice audit pass**. Evidence Match is the authoritative measure — it reports which of the job's requirements the resume actually evidences, with the exact excerpt behind each.
• Legacy ATS/HR scores remain available for continuity but are diagnostic only. No universal ATS score exists across recruiting systems, so never present one as the result or retry a run to chase it.
• Report gaps as "no evidence found in the resume", never as "the candidate lacks this". Absence of evidence is not proof of absence.
• A 75% evidence match with human prose beats a 90% keyword-stuffed draft — stuffing raises keyword counts and lowers evidence quality.

---

NATIVE RESUME TEAM (mandatory control plane)

Only after the candidate-fit preflight passes, read `commands/resume-team.md`, run its host preflight, and invoke `native_resume_team.py` for the current host with the same `run_id`, `case_id`, `as_of_date`, exact private JD file, and a prospective non-existing output path. Do not create the path first. Do not manually reproduce the role sequence for a publishable draft. The runtime enforces:

1. `resume-researcher` receives only the job description. 2. `resume-writer` receives only the master resume and validated Researcher artifact. 3. `resume-auditor` independently audits the exact Writer draft. 4. `resume-editor` is called only after `FAIL`, corrects only named findings, and is followed by a fresh Auditor review. Maximum two corrections.

The coordinator must validate every `resume-team-handoff/v1` digest and lineage, require distinct role identities, and fail closed on malformed, stale, replayed, ambiguous, unavailable, timed-out, or side-effecting results. The coordinator must not write or silently repair the resume itself. Require macOS or Linux; Windows preflight fails closed with `POSIX_RUNTIME_REQUIRED`.

For Codex only, omit model flags by default because the hardened subprocess ignores user configuration and transient parent-session settings; do not claim an inherited profile, model, or Ultra setting. Add `--model <exact-model>` and/or `--reasoning-effort ultra` only when the user explicitly requests those pins. There is no runtime profile option, and Claude must not receive Codex-only pins.

Accept the `resume-team-result/v2` runtime result only