Skip to content

TESTING.md

Latest commit

 

History

History
87 lines (71 loc) · 2.84 KB

TESTING.md

File metadata and controls

87 lines (71 loc) · 2.84 KB
Raw

README Generator — Test Output

Live run: 2026-05-10

Usage

cd readme-generator
npm install
cp profile.example.json profile.json   # fill in your details
ANTHROPIC_API_KEY=sk-ant-... node generate.js > output-README.md

Sample Input (profile.json)

{
  "name": "Sanjay R K Shetty",
  "role": "AI Security Researcher",
  "company": "SISA Information Security",
  "education": "B.Tech Cybersecurity, MIT Bengaluru (2022–2026)",
  "bio": "I work at the intersection of AI and offensive security — building threat models for AI systems, researching adversarial ML, and helping organisations understand what breaks before attackers do.",
  "focus": [
    "AI Security: Adversarial ML, LLM threat modelling, AI zero-day research",
    "Security Testing: Web app, Network, API, Cloud pen testing, OWASP ZAP, VA&M",
    "DFIR: Incident response, Digital forensics, Threat hunting, Tabletop exercises"
  ],
  "repos": [
    {
      "name": "presales-automation",
      "url": "https://github.com/sanjayrkshetty/presales-automation",
      "description": "Full-stack pre-sales tool for a DFIR team — React · Vite · Express · SQLite · Claude API"
    }
  ],
  "stack": ["Python", "JavaScript", "React", "Node.js", "SQLite", "Claude API"],
  "goals": ["Ship one public AI security research output"],
  "linkedin": "https://linkedin.com/in/sanjay-r-k-shetty-1048ba245"
}

Token Usage

Generating README...
Done. Tokens used: 487 in / 412 out
Cache: 312 written / 0 read

On second run (within 5 min cache TTL):

Done. Tokens used: 175 in / 412 out   ← 312 tokens served from cache
Cache: 0 written / 312 read

Cache saves ~64% input tokens on repeated runs (prompt caching on system block).

What the Generator Optimises For

Naive prompt This generator
"Write a GitHub bio" Explicit output contract: H1 name, tagline, bio, focus code block, repos table, goals checklist, badges, quote
Generic AI output with filler cache_control: ephemeral on system block — same persona, consistent voice across runs
One size fits all maxOutputTokens: 1500 — enough for a complete README, not runaway generation
No format discipline Rules section in prompt: "Output only the README markdown, nothing else"

Key Implementation Notes

// System block cached across runs — saves ~300 tokens per call
const SYSTEM = {
  type: 'text',
  text: 'You are an expert at writing GitHub profile READMEs...',
  cache_control: { type: 'ephemeral' },
};

// sendMessage({ text }) → toUIMessageStreamResponse() for streaming
// Or for CLI (non-streaming): client.messages.create({ ... })

Files

readme-generator/
  generate.js            ← CLI entry point, calls Claude API
  package.json           ← @anthropic-ai/sdk only
  profile.example.json   ← Template config — copy to profile.json