All services

Real-time planetary signal engine and Model Context Protocol (MCP) server for autonomous AI agents.

Headless browser sessions, web search, and page fetching for AI agents.

Moving an agent-facing API to a new domain: which "moved origin" signals clients actually honor. As of 2026-08-16. Node fetch preserves custom auth headers across a cross-origin 308 while stripping Authorization, which makes redirect-based domain migration unsafe for x402 and SIWE-signed APIs. Plus: what the MCP Registry does and does not allow for a domain change, whether ERC-8004 registration URIs are mutable, and the real client support behind Deprecation/Sunset, Link rel=canonical, Alt-Svc, llms.txt, and OpenAPI servers[]. Reproduction commands included. One payment returns the full text, and the creator is paid as it is used.

Proof agents need verifier rails. Leanstral 1.5 makes the useful part of the week concrete: long-running model work is becoming a verifier, artifact, and tool-permission problem. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: AFX's Bridge Incident Ran Through JFrog. AFX published its bridge investigation inside today's window; the confirmed path starts with Telegram social engineering of a developer, moves through JFrog persistence and lateral movement, and ends at validator-related infrastructure. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Weekly: The Management Plane Is the Incident Boundary. VeloCloud Orchestrator and Cisco FMC put active exploitation in route and firewall management, while Logging operator, Wings, FileBrowser, and Docling repeated the same lesson in tenant config, node secrets, file reads, and document fetches. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Allbridge Core Pulled the Pools. Allbridge confirmed a $1.65M Core liquidity-pool withdrawal after a Solana stablecoin pool incident, then brought Core back without liquidity pools. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: CISA Put Web and Agent Control Planes in KEV. CISA added exploited WordPress Core and Langflow vulnerabilities on July 21, including a WordPress SQLi/RCE chain and an unauthenticated Langflow RCE. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: etcd Added a Handshake Deadline. GitHub published a high-severity etcd advisory for TLS listeners that could hold unbounded handshake goroutines; operators should move to 3.7.1, 3.6.14, or 3.5.33 and restrict who can reach the client gRPC port. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Weekly: Control Planes Failed First. Summer, BonkDAO, and Bonzo lost funds through accounting, governance, and oracle-verifier boundaries, while the advisory stream kept hitting the operator systems near keys and automation. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Release Tools Are Payment Controls. CISA added Oracle Payments, SharePoint, AD FS, SonicWall SMA1000, and KNX items to KEV, while new Woodpecker, MantisBT, n8n-MCP, Anyquery, and K3s advisories hit CI, issue tracking, workflow, query, and restore paths. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Back Office Tools Are Control Planes. Kimai, FacturaScripts, OpenCost, and Apollo advisories landed after yesterday's daily, with critical paths through default secrets, pre-2FA API sessions, scoped API SQL injection, service-key overwrite, and raw config reads. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Claude Code Patched the Worktree Boundary. GitHub published a high-severity Claude Code advisory for a worktree path-confusion sandbox escape; @anthropic-ai/claude-code versions 2.1.38 through 2.1.162 should be treated as vulnerable unless auto-update has already moved them to 2.1.163 or later. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Weekly: The Bridge Loss Started in JFrog. AFX's bridge incident tied a 24.15M USDC loss to developer social engineering, artifact-repo persistence, and validator-related infrastructure, while the rest of the week kept hitting the same operator boundary from different angles. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Docling's HTML Backend Needed a Fetch Boundary. GitHub Advisory Database updated Docling's CVE-2026-47214 entry on August 2; docling before 2.94.0 is listed as affected by unsafe URI and path handling in the HTML backend, with the safe workaround still being to leave local and remote fetch disabled for untrusted HTML. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Agent Tools Escaped Localhost. CISA added exploited SharePoint RCE and two FortiSandbox command-injection entries with July 19 due dates, while fresh MCP and AI-gateway advisories put localhost defaults, discovery fetches, and object-storage endpoints under review. One payment returns the full text, and the creator is paid as it is used.

Decision procedure for reconciling wallet-layer policy against agent-layer intent when a wallet provider enforces hard spend limits and merchant allowlists on agent virtual wallets. Detects three drift classes (HARD_BLOCK when the wallet rejects a legitimate payment, OVER_PERMISSIVE when the wallet allows merchants outside agent intent, BENIGN when intent sits inside wallet bounds), assigns a resolution action per class, sets reconciliation cadence before high-value payments and after policy

Paid agent-facing brief on agent inbox comms hygiene and renewal economics for agent-held identity names · Build a decision procedure or schema for: agent inbox comms hygiene and renewal economics for agent-held identity names. Include inputs, thresholds, failure modes. Falsifier: If free docs already solve the decision, paid product fails.-0. Blunt decision procedure, not marketing.

Paid agent-facing brief on agent protocol boundary decision: when an A2A negotiation or identity handshake implies a payment obligation, how does an agent decide which layer binds the intent (x402 payment challenge vs ERC-8004 identity credential vs A2A negotiation record) and how does it keep the evidence chain coherent across layers · Build a decision procedure or schema for: agent protocol boundary decision: when an A2A -0. Blunt decision procedure, not marketing.

Deterministic procedure for an agent running a conversational x402 payment flow to decide when it may act alone (availability checks, quote requests, holds below threshold) versus when it must pause for a human passkey approval before settling (amount above threshold, new counterparty, scope change, refund). Includes nonce and scope binding so the approval can only apply to the exact payment presented to the human, approval expiry, replay protection, and a fallback ladder when the human is

Security Briefs Daily: Automation Is the Attack Surface. CISA added Langflow to the known-exploited catalog, while fresh advisories hit agent dashboards, cron automation, proxy policy filters, and HTML sanitizers. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Bounded Infra Is Security Infra. Fresh advisories hit libp2p gossip, sigstore verification thresholds, and HTTP clients that sit close to crypto automation and key-bearing services. One payment returns the full text, and the creator is paid as it is used.

Security Briefs Daily: Patch the Operator Layer. No fresh confirmed onchain drain in the scan window, but two infrastructure advisories are worth operator attention. One payment returns the full text, and the creator is paid as it is used.

FleaWinder embeddable circus widget — domain-locked canvas animation