An Agent Skill for evidence-led resume reviews, job-description tailoring, exact edits, diagnostic scoring, and complete rewrites across IT and cybersecurity roles.
A full review produces an editable Markdown report and a styled PDF generated from the same source. The skill does not rank candidates, predict interviews, or invent missing achievements.
Attach the original resume when possible. A PDF or DOCX lets the agent check extraction and visible layout; pasted text supports content review only.
With a job description:
Use the cyber-resume-reviewer skill to review my attached resume against the attached job description. Give me a candid fit assessment, prioritized findings, and exact edits supported by my resume. Do not invent metrics or experience. Deliver the full report as Markdown and PDF.
Without a job description:
Use the cyber-resume-reviewer skill to review my attached resume for IT and cybersecurity roles. Identify the strongest evidence, the highest-value fixes, and one or two plausible target directions. Mark layout checks you cannot perform as not assessed. Deliver the full report as Markdown and PDF.
For a rewrite:
Use the cyber-resume-reviewer skill to rewrite my resume for a security engineering role. Use only facts in the source. Put questions for stronger claims outside the clean resume.
Tagged releases publish two clean archives on the Releases page:
cyber-resume-reviewer.skillhasSKILL.mdat the archive root for.skilltooling.cyber-resume-reviewer-claude.zipwraps the files in acyber-resume-reviewer/folder for Claude's web uploader.
Neither archive contains repository documentation, tests, workflow files, the changelog, or .gitignore.
| Host | Install and invoke | Support notes |
|---|---|---|
| Codex CLI or IDE | Extract the .skill contents to ~/.agents/skills/cyber-resume-reviewer/, then use $cyber-resume-reviewer or let Codex select it. |
Codex supports local Agent Skills. Standalone and plugin distribution differ by surface. |
| Claude Code | Extract to ~/.claude/skills/cyber-resume-reviewer/ for personal use or .claude/skills/cyber-resume-reviewer/ in a project. Invoke /cyber-resume-reviewer. |
Claude Code follows the Agent Skills format and supports bundled files and scripts. See Claude Code skills. |
| Claude web | Upload cyber-resume-reviewer-claude.zip under Customize > Skills. |
Custom skills require Skills and code execution to be enabled. Account and workspace controls apply. See Claude custom skills. |
| ChatGPT | In the desktop app, open Skills in the sidebar and select an installed skill with @. |
OpenAI documents standalone skills for ChatGPT desktop, but its public guide does not promise direct .skill upload. ChatGPT web and mobile require plugin distribution, which this repository does not provide. See OpenAI skill docs. |
| Gemini CLI | Run gemini skills install ./cyber-resume-reviewer.skill, then /skills reload. |
Gemini CLI accepts local .skill packages. See Gemini CLI skill management. |
| Gemini Spark | Copy or rename the .skill file to .zip, then upload it from Spark's Skills page. |
Spark accepts a ZIP with SKILL.md at its root. Region, subscription, and account limits apply. See Gemini Apps skills. |
The core format follows the Agent Skills specification. Install paths, upload rules, script permissions, and feature availability remain host-specific.
The skill uses four lenses:
| Lens | Intent |
|---|---|
| Machine-read | Check text extraction, reading order, and parser risks without claiming an ATS outcome. |
| Human-skim | Test the first impression, hierarchy, clarity, and positioning. |
| Human-believe | Check whether scope, ownership, technical claims, and results have evidence. |
| Human-act | Align the document with a stated target while preserving requirement wording and unknowns. |
A standard report selects from these sections:
| Section | Intent |
|---|---|
| Assessment | State the target, main conclusion, strongest evidence, and highest-impact repair. |
| Prioritized findings and exact edits | Connect each observation to source text and give a supported action or replacement. |
| Target fit | Separate employer requirements from candidate evidence and reviewer inference. |
| First impression and format | Report checks performed on the source file and mark unavailable checks unassessed. |
| Diagnostic score | Provide an optional editorial score with the selected profile and assessed coverage. |
| Next actions | Order the smallest set of changes that improves the submission. |
- This root
README.mdis the public guide to prompting, installation, portability, dependencies, and releases. cyber-resume-reviewer/README.mdis the maintainer guide to the skill's behavior and local helpers. Release archives exclude it.cyber-resume-reviewer/SKILL.mdcontains the workflow, truth rules, output routing, and quality gate.references/holds role taxonomy, evidence checks, scoring, parser risk, transitions, leadership guidance, and PDF rules.templates/provides report, quick-review, exact-edit, rewrite, interview-story, and change-log structures.scripts/analyze_resume_text.pyreports local text signals without making hiring or ATS claims.scripts/render_report.pyandassets/report.cssturn a Markdown review into the styled PDF.schemas/andscripts/validate_report.pysupport optional JSON output.
The review instructions work without the bundled renderer. Matching the supplied PDF design requires code execution plus:
- Python 3.9+,
PyYAML, andbeautifulsoup4 - Python Markdown or
pandoc - WeasyPrint or
wkhtmltopdf pdftoppm,pdftotext,pdfinfo, andpdffontsfor the full visual check
Install the optional Python set with:
python3 -m pip install -r cyber-resume-reviewer/scripts/report-requirements.txtThe bundled rendering path does not require network access. It writes page images beside the PDF so the agent can inspect every page before delivery.
Pushing a v* tag runs .github/workflows/publish-skill.yml. The workflow builds both archives from a runtime allowlist, records SHA-256 checksums, and attaches the files to the matching GitHub Release. A manual run builds the same files as a workflow artifact for testing.