AOI Prompt Injection Sentinel

S-DNA: AOI-2026-0215-SDNA-PG01

What this is

A lightweight, public-safe prompt-injection detector that scores input text and outputs:

• severity (0–4)
• action (allow/log/warn/block)
• reasons + matched rule ids

What this is NOT

• No webhook, no outbound calls, no auto-posting.
• No secret handling.

Usage

Analyze text (argument)

node skill.js analyze --text="..."

Analyze stdin

echo "..." | node skill.js analyze --stdin=true

Output

JSON to stdout.

Release governance (public)

We publish AOI skills for free and keep improving them. Every release must pass our Security Gate and include an auditable changelog. We do not ship updates that weaken security or licensing clarity. Repeated violations trigger progressive restrictions (warnings → publish pause → archive).

Support

• Issues / bugs / requests: https://github.com/edmonddantesj/aoi-skills/issues
• Please include the skill slug: aoi-prompt-injection-sentinel

Links

• ClawHub: https://clawhub.com/skills/aoi-prompt-injection-sentinel

License

MIT (AOI original).