API Reference

Base URL: https://api.injectshield.dev

Authentication

Pass your API key as a bearer token:

Authorization: Bearer is_live_…

Or via header: X-API-Key: is_live_….

POST /v1/scan

Returns a verdict and (optionally) a sanitized version of the input.

Request body

Field	Type	Default	Description
`text`	string	—	≤ 100,000 chars. Required.
`context`	enum	"unknown"	git_commit · web_content · user_input · file_content · email · tool_output
`options.sensitivity`	enum	"medium"	low · medium · high — threshold tuning
`options.return_cleaned`	bool	true	Return sanitized text with matches replaced.

Response

{
  "safe": false,
  "confidence": 0.94,
  "threat_type": "instruction_injection",
  "patterns_matched": ["ignore-previous", "system-tag-injection"],
  "cleaned_text": "[REDACTED:instruction_injection]",
  "semantic_score": 0.91,
  "semantic_label": "negative",
  "notes": [],
  "request_id": "req_abc123…"
}

POST /v1/keys (signup)

Public, throttled to 5 signups/IP/day. Sends an API key to the supplied email.

curl -X POST https://api.injectshield.dev/v1/keys \
  -H "Content-Type: application/json" \
  -d '{"email":"[email protected]"}'

GET /v1/keys/me

Returns the email, tier, and monthly usage for the bearer key.

GET /v1/usage

Returns current-month request count + blocked count for the bearer key.

POST /v1/checkout

Creates a Stripe Checkout Session. Body: {tier:"hobby"|"team"|"pro", email}. Returns {url}.

POST /v1/demo

Public, IP-throttled. Same shape as /v1/scan minus auth and usage logging. Useful for the live demo on the landing page; not intended for production traffic.

Threat categories

instruction_injection — "ignore previous instructions" and variants
system_override — system-prompt leak attempts, role-tag forgery, ChatML/Llama special tokens
role_hijack — "you are now…", DAN/Developer-Mode personas
exfiltration — request to send data to external URL or markdown image with query string
schema_attack — references to known attack schemas (e.g. OpenClaw)
encoding_smuggle — base64-decoded directive smuggling
invisible_text — zero-width / bidi / Unicode-Tag smuggling
tool_abuse — synthetic tool/function-call directives in untrusted text
jailbreak_classic — DAN, "no restrictions", etc.

Limits + errors

Status	Code	Meaning
400	bad_json / missing_text / bad_email	Malformed request
401	missing_api_key / invalid_api_key	Auth failed
402	monthly_limit	Plan quota exhausted — upgrade
413	text_too_large	Text > 100K chars
429	rate_limited / signup_throttled / demo_rate	Backoff and retry

Defense in depth

InjectShield is one layer of a defense-in-depth strategy. Pair it with: prompt-template constraints, tool argument validation, output filtering, and least-privilege agent permissions. We will not claim 100% detection — that's mathematically impossible.