resume

Shared Resume Workflow

This is the shared, agent-neutral definition of the resume workflow.

Use it when you need to:

• improve an existing resume draft;
• turn project notes or raw material into resume content;
• select a template before intake;
• run structured guided intake for missing fields;
• keep the workflow reusable across Claude and non-Claude agents.

Core rule

This installed skill bundle is self-contained.

When the skill is installed correctly, this directory should contain:

• template_catalog_cli.py
• template_store_cli.py
• agent_intake_cli.py
• host_cli.py
• render_cli.py
• resume_runtime/
• resume_core/examples/

Treat the bundled resume_runtime/ under this skill directory as the primary shared interface.

Use these public entrypoints:

• template_catalog_cli.py — list built-in and stored templates and derive template_context
• template_store_cli.py — save personal templates and promote candidate templates
• agent_intake_cli.py — host-facing outer intake entrypoint
• host_cli.py — lower-level structured session control
• render_cli.py — render markdown/html/css output bundles from a manifest and profile

Equivalent bundled module paths also exist under:

• resume_runtime/template_catalog_cli.py
• resume_runtime/template_store_cli.py
• resume_runtime/agent_intake_cli.py
• resume_runtime/host_cli.py
• resume_runtime/render_cli.py

Claude-specific wrappers under .claude/skills/resume/ are compatibility adapters, not the primary shared contract.

Recommended workflow

1. Choose the template first

Before parsing materials or asking for missing information, choose a template.

• In the installed skill bundle, built-in template manifests live under resume_core/examples/templates/
• Built-in template assets, including CSS, live under resume_core/examples/template-assets/
• The current built-in templates are typora-classic and markdown-basic
• load built-in and stored templates through template_catalog_cli.py
• read asset_paths from the catalog CLI response when you need the concrete markdown/html/css file locations for a template
• show template cards with template id, style, use cases, and required-content summary
• let the user choose a built-in template, a stored template, upload a new template, or ask for a derivative template
• once a template is chosen, use the returned manifest + derived checklist as the template_context for agent_intake_cli.py

If you know the absolute path of this installed skill directory, enumerate built-in templates with:

python3 /path/to/installed/skill/template_catalog_cli.py

If you are already in the installed skill directory, you can also run:

python3 template_catalog_cli.py

If you are developing inside this repository root, the bundled entrypoint is:

python3 skills/resume/template_catalog_cli.py

You can still override the defaults when needed:

python3 /path/to/installed/skill/template_catalog_cli.py \
  --examples-root /path/to/installed/skill/resume_core/examples \
  --generated-at 2026-04-16T12:00:00Z

If you need to inspect a built-in template directly, start from:

• resume_core/examples/templates/typora-classic.v1.json
• resume_core/examples/templates/markdown-basic.v1.json
• resume_core/examples/template-assets/typora-classic/style.css
• resume_core/examples/template-assets/markdown-basic/style.css

Do not assume the template files live beside skills/resume/SKILL.md. The skill file defines the workflow; the built-in template manifests and assets live in the repository paths above.

2. Choose the content path after template selection

Use one of these paths:

• direct content entry
• parsing existing materials
• follow-up questioning only for missing template-required fields

If the work is straightforward rewrite/polish, stay in direct editing. If the work depends on structured intake state, missing-field projection, or follow-up rounds, route through the public intake/session CLIs.

3. Use the public intake flow for structured collection

For structured intake and follow-up:

• call resume_runtime/agent_intake_cli.py as the public outer entrypoint
• provide template_context when a template has already been chosen
• let the runtime decide whether the turn should stay freeform, ask for material, parse material, continue a structured session, or hand off to drafting
• only ask the current batch of questions materialized by the runtime
• at the recommended-only boundary, collect only a yes/no continuation decision

4. Draft against the selected template

When enough information is available:

• produce the resume draft that matches the selected template’s required structure
• localize section headings, field labels, and fixed template copy to match the user’s language; Chinese requests should render Chinese labels by default
• preserve evidence-based wording and truthful ownership
• list any remaining missing facts explicitly instead of guessing

5. Render final template artifacts through the public renderer

When a profile is ready:

• use resume_runtime/render_cli.py or the shared rendering helpers
• generate markdown/html/css outputs from the selected manifest and profile
• treat these rendered assets as the portable output bundle for host integrations and Typora-first preview flows

Host-neutral request shape

When calling resume_runtime/agent_intake_cli.py, use the public request envelope documented in README.md and this shared skill.

At minimum, expect these concepts:

• version
• turn
• optional template_context
• optional materials
• optional drafting_started

Output expectations

A good run should usually produce:

• a selected template or template decision
• a filled or partially filled profile
• a clear missing-information list when facts are still absent
• a resume draft or rendered output bundle when enough information is present

Boundary rule

Shared workflow logic belongs in resume_runtime/ and shared docs like this file. Agent-specific prompt shaping belongs in that agent’s adapter layer.