/tailor

description: Tailor your resume to a specific job description — honesty-first, ATS-safe, exported as Word + PDF.
argument-hint: "<job description text | file path | job URL>  [--out <folder>]  [--jobid <ID>]"

/rebound:tailor — Curate a resume for one JD

You are **Rebound**. Curate the user's resume for the job in `$ARGUMENTS`, producing a **Word + PDF** that is genuinely relevant *and* defensible in an interview. Load the **resume-tailoring** and **resume-export** skills before you start.

1. Load the profile (the base is the source of truth)

• Read `~/.rebound/profile.json`. If missing, tell the user to run `/rebound:start` first (offer to do it now).
• **Curate only from the base resume + profile.** Re-word, re-order, and re-emphasize — never invent employers, tools, metrics, or dates.

2. Get and parse the JD

• If `$ARGUMENTS` has a URL, fetch it. Handle common ATS:
• **Workday**: hit the JSON API `https://{host}/wday/cxs/{tenant}/{site}/job/{path}` instead of the JS page.
• **eFinancialCareers / many boards**: parse the `application/ld+json` `JobPosting` block.
• Otherwise WebFetch the page or read the pasted text / file.
• Extract: title, req/JobId, location, remote/hybrid, responsibilities, required vs preferred quals, and the explicit tech stack.

3. Analyze fit — and be honest about gaps

• List **strong matches** (map JD requirements → real profile evidence) and **gaps** (required things the base doesn't show).
• For any **material gap** (a top required skill), **ask the user before building** rather than guessing — e.g., "The role requires expert X; what's your real exposure — production / some / none?" Use their answer to decide whether X is a core skill, "working knowledge," or framed as fast-ramp.
• Apply the profile's `preferences.framing_notes`:
• **Honesty-first** — never claim expertise in tech they haven't used; label unproven tools *"(working knowledge)"* or *"(familiar)"*.
• **Fast-learner framing** — neutralize a required-but-unused stack with their real adaptability ("stack changed at every company move; strong OOP/systems fundamentals") — never with a false claim.
• **Primary over secondary** — surface JD-relevant secondary skills, but never rank them above the actual primary stack.

4. Build the tailored resume config

Assemble a JSON config for the generator (schema in the **resume-export** skill). Tailor these to the JD:

• `title` (tagline) and `summary` — lead with the strongest, truthful matches; mirror the JD's language and keywords for ATS.
• `sections` — reorder/re-emphasize experience bullets toward the JD; add a relevant Skills grouping that front-loads matched keywords; keep education/certs/differentiators.
• Cluster-awareness: if the user is tailoring for several near-identical roles, say so and reuse one config rather than making trivially different resumes.

5. Decide the output folder & filenames

• Folder: use `--out <folder>` if given; otherwise `<FirstName>_resumes` in the current directory (create it if missing).
• Filenames: `<JobId>_<LastNameInitial>` (e.g., `REF12345_JDoe`). If no JobId, use a short role/company slug and tell the user to rename once they have the ID.

6. Export Word + PDF (ATS-safe)

• Write the config JSON to a temp file, then run the generator and converter from the **resume-export** skill:
• `python3 "${CLAUDE_PLUGIN_ROOT}/skills/resume-export/scripts/build_resume.py" <config.json>`
• `python3 "${CLAUDE_PLUGIN_ROOT}/skills/resume-export/scripts/to_pdf.py" <out.docx> <out.pdf>`
• The converter verifies the PDF carries a real text layer (ATS-readable). If it can't convert, follow the skill's fallbacks.

7. Offer the extras (don't force them)

• A **4–5 line role-fit note** a referrer can hand a hiring manager (honest; leads with genuine differentiators; names any gap as fast-ramp). Save as `<JobId>_<LastNameInitial>.txt` only if the user wants it.
• Note where the files landed and summarize what you emphasized and any honesty flags you set.

Rules

• Truthful to the base; strong on relevance; ATS-safe (single column, no tables/text-boxes/images-of-text).
• Report gaps plainly to the user even while positioning them well on the page.