skip to content
fizzgig
sign in
// product
code auditdetect what shippedbrand and reachsoonlift qualityperfect memorysoondecision graph
// build better
sourdough startersproject kicks-off in 1 prompt
// resources
pricingdocsblogabout
sign inrequest access →
audit suite/security/prompt_injection_scan
securityv0.2.0pro · $0.002/call

fizzgig__prompt_injection_scan

flags prompt content for injection-vector smells.

// ai does this over time

your app fetches user-controlled content (an email body, a scraped page, a tool response) and forwards it into a prompt. somewhere in that content is IGNORE PREVIOUS INSTRUCTIONS. now your assistant is doing whatever the attacker asked.

// the tool does

6 categories with content-type-aware severity: system-prompt overrides, role breakouts, jailbreak personas, prompt-extraction attempts, capability exploits, indirect-injection payloads. Weighs each by content_type (user_input, fetched_content, prompt_template, tool_response) so risk grades scale with where the content came from.

// so you can

treat fetched content as adversarial. the AI didn't, and won't, unless you check.

● live● v0.2.0● pro
// input schema
schema · application/json
{
"type": "object"
"required": [
"project"
]
"properties": {
"project": {
"type": "string"
"description": "the project slug or path"
}
"strict": {
"type": "boolean"
"default": false
"description": "fail on warnings, not just errors"
}
}
}
// output schema
schema · application/json
{
"type": "object"
"properties": {
"ok": {
"type": "boolean"
}
"findings": {
"type": "array"
"items": {
"type": "object"
"properties": {
"severity": {
"enum": [
"info"
"warn"
"high"
"critical"
]
}
"message": {
"type": "string"
}
"fix": {
"type": "string"
}
}
}
}
}
}
// example call from cursor
~/myapp — example output
→ fizzgig__prompt_injection_scan(project="myapp")
{
"ok": false,
"findings": [
{ "severity": "high",
"message": "policy uses user_id without auth.uid()",
"fix": "USING (auth.uid() = user_id)" }
],
"scanned": 3, "duration_ms": 142
}
// reviews
@maya.codes★★★★★
2 days ago

caught a policy that would have leaked every user's comments. shipped a fix in 4 minutes.

@solo_at_3am★★★★★
1 week ago

first tool i installed. it's the one that pays for itself.

@vibebuilder★★★★☆
2 weeks ago

works great. one false positive on a join table — easy to ignore.

// primary action
add to your editor
paste this into your mcp config.
.cursor/mcp.json
{
  "fizzgig": {
    "url": "https://mcp.fizzgig.ai",
    "tools": ["prompt_injection_scan"]
  }
}
full setup guide →
// pricing
$0.002 / call
included free in pro plan.
// related tools
secret_leak_finder
v0.9.0
→
rls_checker
v0.6.0
→
env_auditor
v0.4.0
→
fizzgig

the fluffy guardian of vibe-coded apps. growls at insecure code so you don't have to.

all systems operational

product

code auditbrand and reach (soon)perfect memory (soon)pricingdocschangelog

community

sourdough startersdiscord (soon)github (soon)x / twitter (soon)rss (soon)

company

aboutcontactterms (soon)privacy
© 2026 fizzgig labs.
v1.4.2 · 2026-05-01