Demo User
ai_engineerentryClientX_AI_Hiring
Sales Investigation · FMCG Data Assessment
68/ 1004 must-haves met · 1 gap · 0 unassessed · 23 evidence items
MUST-HAVE · HIGH WEIGHTPrompt Engineering Discipline71%
For the hiring manager
Writes prompts that produce the correct general output. You do not expose tool/function definitions to the LLM; the system uses plain prompts instead of tool-calling, so tool design is weak.
Observations
- You do not expose tool/function definitions to the LLM; the system uses plain prompts instead of tool-calling, so tool design is weak.
- Your context management is adequate: the schema is assembled dynamically with table definitions and sample rows, but includes all tables/views without pruning; no explicit token budgeting is evident.
Evidence
repo:src/agent_graph.pybuild_schema_text() builds full schema string with all tables and views; no token limit or pruning applied.repo:src/entity_validator.pyEntity validation uses LLM extraction then database lookup; context is sparse but focused on entities.
GAP · LOW WEIGHTAI-Augmented Design Process33%
For the hiring manager
Uses AI to generate starter code or expand on an idea. Your prompts are mostly directive (e.g., 'read all files', 'add security', 'fix SQL', 'update MD files') with minimal evidence of original analytical framing or challenging the AI's suggestions.
Observations
- Your prompts are mostly directive (e.g., 'read all files', 'add security', 'fix SQL', 'update MD files') with minimal evidence of original analytical framing or challenging the AI's suggestions.
Evidence
- Candidate's verbatim instructions are generic: 'Go through everything, ask clarifying questions', 'Read ALL the text files', 'add protections', 'Fix this SQL query', 'update the md files', 'redo the reconciliation.md file', 'restructure approach.md'. No evidence of candidate challenging the AI's analytical choices.