Prompt-injection Guard checks and public AI red-teaming arena data from Heisto on Solana.
Scored across 5 tools
Each tool targets a distinct resource or purpose: config, a single sentinel, a sentinel list, settled verdicts, and a text filter. The closest pair is get_sentinel vs list_sentinels, but the standard verb/noun split keeps them clearly separable. guard_check is orthogonal to the arena read surface but its purpose is unambiguous.
Three tools use verb_noun (get_sentinel, list_sentinels, guard_check) while two use bare noun compounds (arena_config, recent_attempts), so read operations aren't signaled by a consistent pattern. Names are still readable and snake_case throughout, but the convention is mixed.
Five tools is well-scoped for a public read surface plus one safety filter, with no redundant entries. It sits at the lean end, and a single detail-read tool for attempts would round it out.
The read side covers config, sentinel listing/lookup, and settled verdicts, but there is no way to fetch a single attempt in detail or filter attempts by sentinel, and no submission path if arena participation is in scope. Agents can read the public state but hit dead ends on per-record drill-down.
Read public Mainnet pool limits, fee split, breach rule and refund delay.
Inputs: no named parameters.
Read a public Sentinel objective and public instructions. Never private instructions.
Inputs: id (required).
Check text for prompt-injection patterns. A first filter, not a safety guarantee.
Inputs: text (required).
List public Mainnet Sentinels or unfunded practice templates.
Inputs: network.
Read settled public verdicts with shortened wallets. Never raw prompts.
Inputs: sentinelId (required), limit.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Check user prompts for injection/jailbreaks; your own guard learns from your labels. Works keyless.
Prompt injection detection API for AI agents. Scan untrusted text before passing it to an LLM.
Prompt injection and system-prompt extraction testing for LLM agents. Free 8-probe sample.
Solana token safety for AI agents — rug-pull, honeypot & Token-2022 trap detection before you buy.
Add one secure layer between your agents and this server.
