Security

Bind a memory record to source identity, observation time and content hash

Validate tool output against a bounded top-level JSON Schema subset

Verify active task lease ownership before execution or mutation

Verify hash-linked session identity and monotonic sequence continuity

Enforce delegated credential scope, requester and expiry without receiving raw credentials

Reject x402 quotes that cannot remain valid through authorization and retry

Bind verified OAuth issuer and audience claims to the intended receiver

Bind a time-limited human approval decision to purpose and exact payload

Verify cancellation state and unresolved side effects

Bind task state, sequence and checkpoint hash in signed evidence

Validate durable, isolated and replay-safe agent task envelopes

Capture supplied public explorer or RPC evidence with immutable content hashes

Bundle independent source observations while exposing agreement and disagreement

Test whether source observations meet an explicit maximum-age policy

Normalize public text, HTML, JSON or XML documents into source-backed JSON

Canonical PHION catalog snapshot or digest-based delta with exact execution templates and signed evidence

Preserve attributable public social-source observations and explicit identity limitations

Search supplied public sources and return only literal source-backed matches

Institutional 13F whale filings, hedge fund shifts, and audited EDGAR risk intelligence.

Base token verdict for trading agents: risk level, score and flags (proxy, mint, pause, blacklist, owner, DEX liquidity on Uniswap v2/v3 and Aerodrome), each with the on…

Analyze French or English marketing and interface copy. Returns a clarity score, repetition data, sentence metrics, CTA detection and concrete fixes.

NOT AN AUDIT — heuristic screen: contract safety screen — verification status, proxy and owner-privilege heuristics, holder concentration, and a plain-English risk verdi…

NOT AN AUDIT — heuristic checklist, not a penetration test: prompt red-team screen ($0.10) — score a system prompt against 7 prompt-injection weakness checks, each with…

NOT AN AUDIT — heuristic pattern sweep, not a security audit: leaked-secret sweep ($0.02) — scan a URL or pasted text for exposed AWS keys, GitHub tokens, private keys,…