A Flask-based CV/resume builder and generator that produces ATS-friendly DOCX/PDF files and visual PDF previews with multi-language support, photo cropping, company logo fetching, and customizable themes.
CV Forge has 4 tools with basic descriptions and partial schemas. Tool names are action-oriented (fetch-, generate-) but descriptions lack depth on WHEN to use each tool and what distinguishes them. Input schemas are present but incomplete: 'personal' and 'html' parameters lack type definitions (object structure not specified), and no output schemas are documented. Error handling is minimal, responses are generic HTTP status codes without recovery guidance. No parameter validation rules, enums, or constraints visible. Security is reasonable (SSRF checks, rate limiting, CORS) but tool definitions themselves are sparse.
Fetch logo from a URL domain via Google/Clearbit (SSRF-safe proxy).
Generate ATS PDF from DOCX via LibreOffice.
Generate ATS DOCX from CV data sent in request body.
Generate Pretty PDF from HTML preview content.
Input schemas lack type definitions for complex parameters. 'personal' and 'html' are declared as objects but their internal structure is not specified. LLMs cannot infer what fields belong inside 'personal' (name, email, phone, etc.) or validate before calling.
No output schemas documented. Tools return files (DOCX, PDF) but LLMs have no visibility into response structure, success indicators, or error details. Agents cannot plan downstream actions or handle failures gracefully.
Tool descriptions are under 100 characters and lack context on WHEN to use each tool. 'generate-pdf' vs 'generate-ats-pdf' distinction is unclear, both generate PDFs. Descriptions do not explain the use case difference (pretty vs ATS-optimized) or when to choose one over the other.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
Error responses are generic HTTP status codes (400, 404, 500) without actionable recovery guidance. 'Invalid request' or 'Document generation failed' tells the LLM nothing about what went wrong or what to try next.
No parameter descriptions visible in schema. The 'url' parameter in fetch-logo lacks guidance on format (domain only vs full URL). The 'cv_data' parameter in generate-pdf is marked optional but its purpose and structure are undocumented.