CV Skill

Create professional Harvard-style resumes and CVs from user-provided candidate descriptions, structured data, or existing resumes. Use when the user wants a...

Security Scan

The skill is internally consistent with its stated purpose (generating Harvard-style resumes from structured input), requests no secrets or unrelated access, and its included script only reads local JSON and writes DOCX/PDF files — nothing appears malicious or disproportionate.

Purpose & Capability

Name, description, SKILL.md, reference docs, example JSON, and the Python generator are all aligned: the package is a local resume generator. No unrelated credentials, binaries, or config paths are requested.

Instruction Scope

Runtime instructions tell the agent to use the included script and schema, review outputs, and avoid hardcoding secrets. The script only reads a user-provided JSON, uses local fonts/styles, writes DOCX output, and optionally calls LibreOffice to convert to PDF. There are no instructions to read unrelated system files, environment variables, or to send data to external endpoints.

Install Mechanism

There is no install spec (instruction-only + local script). The script requires Python and the python-docx package (not declared in metadata) and optionally LibreOffice/soffice for PDF conversion. This is normal but users should be aware they must install python-docx and have Python available.

Credentials

The skill declares no required environment variables, no credentials, and the code does not read env vars. No sensitive or unrelated secrets are requested.

Persistence & Privilege

The skill is not always-enabled and does not modify other skills or system-wide agent settings. It runs locally when invoked and writes output files to a user-specified directory only.

Assessment

This skill appears to do only what it says: generate DOCX (and optionally PDF) resumes from a JSON input. Before running, verify you have Python and the python-docx package installed, run the script in a directory you control, and inspect generated files before sharing. Because the script will write files, avoid running it with elevated privileges and don't place real candidate PII in the skill folder—use redacted example files. If you need stronger assurance, open and review scripts/generate_resume.py yourself or run it in an isolated environment. If you plan to enable autonomous agent invocation, remember the agent could run this script and write output without additional prompts — only enable that if you trust the skill and its inputs.

Create role-targeted, black-and-white, Harvard-style resumes from candidate descriptions, structured input, or existing resumes.

Use this skill when

• The user wants a professional resume or CV in .docx
• The user gives a rough candidate description and wants the agent to draft the resume from scratch
• The user wants one candidate rewritten into multiple job-targeted versions
• The user provides a PDF, notes, or rough bullets and wants a polished resume
• The user wants tighter, more professional bullets without fluff
• The user wants a Harvard-style layout with larger spacing and clean hierarchy
• The user needs output in a language other than Chinese or English

Workflow

1. Gather candidate data

Use the structured schema in references/input-schema.md.

If you are starting from an existing resume, extract:

• contact info
• summary / positioning
• education
• work experience
• projects
• campus or extracurricular items
• tools, languages, certificates
• target job directions

2. Define track-specific positioning

For each job direction, rewrite:

• resume title
• 2-3 sentence summary
• bullet emphasis within experience
• skills ordering

Keep facts intact. Do not invent results or responsibilities.

3. Generate the resume

Run:

python3 scripts/generate_resume.py --input assets/example_profile.json --track all --output-dir /tmp/cv-output

Generate a specific track:

python3 scripts/generate_resume.py --input candidate.json --track operations --output-dir /tmp/cv-output

Try PDF export when LibreOffice is installed:

python3 scripts/generate_resume.py --input candidate.json --track all --output-dir /tmp/cv-output --pdf

4. Validate before delivery

Check that:

• no hardcoded personal info from unrelated candidates remains
• dates and headings are consistent
• bullets are role-targeted rather than generic
• low-signal items are removed or pushed down
• generated filenames are generic and safe

Layout rules

• Single column
• Black and white only
• Section headers with strong hierarchy
• Larger spacing than default Word exports
• Short, factual bullets
• Avoid self-evaluation phrases such as “责任心强” or “结果导向”
• Prefer evidence and scope over adjectives

Safety rules

• Do not hardcode real candidate data into scripts
• Do not store secrets, API keys, tokens, or .env files in the skill folder
• Keep outputs outside the skill folder unless the user explicitly wants examples saved there
• Use assets/example_profile.json only as a redacted example

Files

• scripts/generate_resume.py: generic generator
• references/input-schema.md: input contract
• references/rewriting-guide.md: track-specific rewriting guidance
• assets/example_profile.json: safe sample input
• agents/openai.yaml: UI metadata