create-skill-card

name	create-skill-card
description	Create a Scoutica Skill Card from candidate documents. Use when asked to generate a professional profile, build a skill card, or convert a CV/resume into a machine-readable format. Reads documents from a folder and outputs 4 structured files (profile.json, rules.yaml, evidence.json, SKILL.md).

create-skill-card

Generate a complete Scoutica Skill Card from candidate documents (CVs, portfolios, certifications).

When to Use

• User asks you to "create my card" or "generate my profile"
• User wants to convert a CV/resume into Scoutica format
• User drops documents and asks for a professional profile
• User says "scoutica scan" or references card generation

Quick Method: Use the CLI

# Auto-detect local AI CLI and generate
scoutica scan <docs-folder> --output <card-folder>

# Or with a specific provider
scoutica scan <docs-folder> --with gemini
scoutica scan <docs-folder> --with claude
scoutica scan <docs-folder> --with ollama
scoutica scan <docs-folder> --with ail

Manual Method: Read Docs and Generate Directly

If you are the AI agent being asked to generate the card, follow these steps:

Step 1: Read All Documents

Read every supported file in the user's folder:

• Text: .md, .txt, .json, .yaml, .csv, .html
• PDFs: extract text via pdftotext or python3 -c "import PyPDF2; ..."
• DOCX: extract via textutil (macOS) or python-docx

Step 2: Generate profile.json

{
  "schema_version": "0.1.0",
  "name": "Alice Developer",
  "title": "Professional Title",
  "seniority": "senior",
  "years_experience": 10,
  "availability": "in_2_weeks",
  "primary_domains": ["Backend Engineering", "DevOps"],
  "skills": ["Python", "Go", "TypeScript", "Django", "FastAPI", "React"],
  "tools_and_platforms": ["Kubernetes", "AWS", "Docker", "PostgreSQL"],
  "certifications_and_licenses": ["AWS Solutions Architect"],
  "specializations": ["Technical Leadership", "System Design"],
  "spoken_languages": [
    {"language": "English", "proficiency": "native"},
    {"language": "German", "proficiency": "fluent"}
  ]
}

Rules:

• NEVER invent skills not found in the documents
• Use standardized names (e.g., "Kubernetes" not "K8s")
• Map seniority from context: "Lead" → "lead", "Manager" → "manager"

Step 3: Generate rules.yaml

schema_version: "0.1.0"

engagement:
  allowed_types:              # permanent | contract | fractional | advisory
    - permanent
    - contract
  compensation:
    minimum_base_eur:         # minimum base per engagement type; omit a type — or the
      permanent: 80000        # whole compensation block — if the documents don't state it
      contract: 600

remote:
  policy: "remote_only"       # e.g. remote_only | hybrid | onsite | flexible
  hybrid_locations: []        # cities/regions, only when policy is hybrid

filters:
  blocked_industries:         # only if the candidate explicitly names them
    - "gambling"
    - "weapons"
  stack_keywords:
    preferred:                # technologies the candidate wants to work with
      - "Python"
      - "TypeScript"
  soft_reject:
    weak_stack_overlap_below: 3   # flag roles matching fewer than N preferred keywords

privacy:
  zone_1_public: [title, seniority, primary_domains, availability]
  zone_2_paid: [, , ]
   [, , ]

Rules:

• Compensation nests under engagement: engagement.compensation.minimum_base_eur.<type>. If salary is not in the documents, omit minimum_base_eur (or the whole compensation block) — never invent a figure.
• Default remote.policy to "flexible" if not stated.
• Only add filters.blocked_industries if explicitly mentioned.
• Rejection rules live under filters (blocked_industries, soft_reject.weak_stack_overlap_below) — there is no top-level auto_reject. This shape is validated by scoutica validate and read directly by the scorer.

Step 4: Generate evidence.json

{
  "items": [
    {
      "type": "github_repo",
      "title": "Project Name",
      "url": "https://github.com/user/repo",
      "skills": ["Python", "FastAPI"],
      "description": "Brief description"
    },
    {
      "type": "certification",
      "title": "AWS Solutions Architect",
      "url": "https://verify.link",
      "skills": ["AWS"],
      "issued": "2024"
    }
  ]
}

Rules:

• Only include URLs found in the documents
• NEVER fabricate links
• Types: github_repo, website, certification, article, talk, portfolio

Step 5: Generate SKILL.md

Use the template in protocol/templates/SKILL.template.md or generate:

---
name: scoutica
description: <Name> — AI-readable professional profile
metadata:
  author: <Name>
  version: 0.1.0
---

# Scoutica

Professional profile for **<Name>** — <Title>.
[... standard structure ...]

Step 6: Copy Rule Templates

Copy evaluation rules from protocol/templates/rules/:

• evaluate-fit.md
• negotiate-terms.md
• verify-evidence.md
• request-interview.md

Step 7: Validate

scoutica validate <card-folder>

Output Structure

<card-folder>/
├── profile.json
├── rules.yaml
├── evidence.json
├── SKILL.md
└── rules/
    ├── evaluate-fit.md
    ├── negotiate-terms.md
    ├── verify-evidence.md
    └── request-interview.md

Schema Reference

The candidate card has no standalone JSON Schema file. Validate the generated card with scoutica validate <card-folder>, which runs tools/validate_card.py against the canonical card files.