Resume Tailoring for Codex

This is a Codex-native adaptation of the external varunr89/resume-tailoring-skill. The upstream skill and supporting references are preserved under references/upstream/.

Source And Provenance

• Upstream repo: https://github.com/varunr89/resume-tailoring-skill
• Installed upstream commit: 9a4a0f20f5983d1b533627b8c5191acd1ca0cd89
• License: MIT, preserved at references/upstream/LICENSE
• Upstream skill: references/upstream/SKILL.upstream.md

When To Use

Use this skill when the user provides, or wants to prepare for, a specific opportunity and has existing resume/CV material, project notes, papers, or experience records that can be truthfully reused.

Prefer claude-resume-kit when the central task is extracting research papers, reports, or codebases into a provenance-aware academic CV knowledge base. Use this skill when the central task is matching a resume/CV library to one or more target opportunities.

Required Inputs

• Target JD, role description, professor/RA opportunity, or opportunity notes.
• Resume/CV library path, existing CV file, or enough raw experience material to build a small library.
• Desired output format: Markdown by default; DOCX/PDF/LaTeX only when the local toolchain or user requirement supports it.

Reference Routing

Read these files only when needed:

• references/upstream/SKILL.upstream.md: full upstream workflow.
• references/upstream/research-prompts.md: company, role, and JD research templates.
• references/upstream/matching-strategies.md: confidence scoring and truthful reframing.
• references/upstream/branching-questions.md: conversational experience discovery.
• references/upstream/multi-job-workflow.md: batch mode for 3-5 similar opportunities.

Workflow

1. Build or locate the experience library.

   • Use resumes/ in the current project by default when it exists.
   • Accept Markdown, plain text, pasted CVs, project notes, or user-provided experience records.
   • Extract roles, dates, companies/institutions, bullets, skills, education, publications, projects, and metrics.

2. Analyze the target opportunity.

   • Parse must-have requirements, nice-to-have requirements, implied skills, keywords, audience, and risk factors.
   • For current or specific opportunities, research public official sources when useful and available.
   • Keep official facts, inferred role needs, and anecdotal context separate.

3. Propose a structure before generation.

   • Recommend section order, role consolidation, title framing, skills grouping, and bullet allocation.
   • Ask for confirmation when the structure materially changes the user's positioning.

4. Match evidence to requirements.

   • Use the upstream confidence model: direct, transferable, adjacent, and impact alignment.
   • Show low-confidence matches and gaps rather than hiding them.
   • For each important reframing, preserve a before/after note and why it remains truthful.

5. Run branching experience discovery when gaps matter.

   • Ask focused follow-up questions instead of a static questionnaire.
   • Capture only experiences the user confirms.
   • Tag new evidence with scope, date, source, and confidence.

6. Generate the tailored output.

   • Produce a tailored resume/CV in Markdown unless another format is requested and feasible.
   • Include a generation report with requirement coverage, evidence mapping, gaps, reframings, and interview-prep notes.
   • If generating files, keep source data and output in the user's chosen project directory.

7. Offer library update.

   • Ask before adding generated or newly discovered content back into the reusable library.
   • Do not overwrite the user's master resume/CV without explicit confirmation.

Truthfulness Contract

• Never invent roles, companies, institutions, dates, publications, skills, tools, metrics, awards, degrees, or seniority.
• Never add a keyword unless it is directly evidenced, adjacent and clearly framed, or explicitly user-confirmed.
• Never inflate "assisted", "contributed", or "implemented a component" into "led", "owned", or "architected" without evidence.
• Keep gaps honest; recommend cover-letter framing, interview preparation, or future work rather than padding the resume.
• For academic CVs, distinguish publications, manuscripts under review, works in progress, course projects, reproductions, and internal reports.

Output Shape

For substantive tasks, return or write:

• Tailored resume/CV draft.
• Evidence mapping table.
• Gap and risk notes.
• Change log or before/after reframing notes.
• Suggested next revision.

For quick critique tasks, lead with the highest-risk issues and then concise fixes.

Boundaries

• Do not automate job applications, recruiter messages, platform submissions, or login-gated actions.
• Do not store secrets, cookies, tokens, private account data, or unrevised sensitive personal information.
• Do not depend on an external API unless the user explicitly requests and configures it.