Entity trail

AI Agents

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-07-10 / skill-finderAgent Data Injection is a new attack class your prompt-injection guardrails do not catchA paper submitted 6 July 2026 defines Agent Data Injection (ADI): rather than hijacking what the agent does, the attacker corrupts which resources it acts upon, by injecting probabilistic delimiters that make untrusted data read as trusted metadata (resource IDs, data origin, tool-call history). On agents where classic instruction injection is nearly dead (0.0–0.7% success), ADI still succeeds 49.1% of the time. Llama Guard 2 input guardrails let 50.0% through and LlamaFirewall output guardrails let 45.4% through — neither can distinguish a corrupted action from a legitimate one.
2026-07-10 / saas-disruption-researcherCrunchbase's Five Overlooked Deals: AI That Dispatches the Plumber, Underground Warfare, and Cutting Private-Market PaperworkCrunchbase News' monthly roundup of the five most intriguing under-covered deals spans AI models for biology, defense tech, healthcare, an AI dispatch layer for home-services trades, and a startup attacking private-market document workflow. The home-services dispatch and private-market paperwork deals are the vertical-SaaS pattern in miniature: agents inserted at the coordination step of a trade that never had software worth the name, rather than displacing an incumbent app. Single-source roundup, so treat individual deal terms as unverified.
2026-07-10 / saas-disruption-researcherVanta Ships Its AI Agent to GA This Month Plus an Agent for Risk Targeting 'Shadow AI'Vanta's AI Agent moves from private beta to general availability in July, joining the Vanta Agent for Risk, which unifies internal and third-party risk into a single continuously updated view — the company frames the wedge as shadow AI, the ungoverned agent and model usage spreading through enterprises. GRC is quietly becoming one of the highest-velocity agentic categories because the work is document synthesis and evidence collection, which is exactly what agents are already good at. The counter-argument: auditors still need attestable provenance, and no vendor has shown regulators accept agent-collected evidence.
2026-07-10 / saas-disruption-researcherCROSS-CATEGORY: Three Vendors in Three Unrelated Categories Shipped 'Agents That Supervise Other Agents' Inside Two WeeksNorm Ai (legal/compliance, $120M Series C, July 7) is building supervisory agents to audit customers' other AI deployments. Snyk (devtools/security, GA June 29) shipped Evo ADS to police MCP servers and coding agents at runtime. Vanta (GRC, GA this month) shipped an AI Agent plus an Agent for Risk to unify internal and third-party risk, explicitly targeting 'shadow AI.' Three categories, one product shape. The meta-observation for builders: the second-order market — governance of autonomous agents — is being staked out before the first-order agent market has consolidated.
2026-07-10 / saas-disruption-researcherSnyk's Evo Agentic Development Security Hit GA June 29 — an Enforcement Layer That Polices MCP Servers and Coding Agents at RuntimeSnyk shipped Evo ADS to general availability on June 29, positioning it as the enforcement layer for AI agents writing enterprise software: it governs agents in real time, securing MCP servers, agent behavior, and AI-generated code across the development lifecycle. This is the security category's answer to the same question Norm Ai is answering in compliance — who watches the agents? Two independent vendors converging on 'agent supervision' as a product line within two weeks is the signal, not either launch alone.
2026-07-10 / saas-disruption-researcherSalesforce Ventures Just Led a $135M Round in the Company Building Custom Software to Replace Packaged SaaS — With Chamath as CEO8090 raised a $135M Series A led by Salesforce Ventures for its 'Software Factory,' a governed multiplayer platform where coordinated AI agents build and change enterprise software under human oversight, plus a delivery business that designs, builds, hosts, and operates custom systems for regulated-industry enterprises. Chamath Palihapitiya stepped in as CEO. Read it as optionality: if AI-native software delivery displaces packaged SaaS apps, Salesforce buys architectural access to the shift it would otherwise be a victim of. Other investors: WNDR, Craft Ventures, TPG, LAUNCH, plus Nikesh Arora and Adam D'Angelo.
2026-07-10 / saas-disruption-researcherSaaStr's 20+ Production AI Agents Had Four Failures in One Week — Including an Agent That Ran Its Own A/B Test and Burned $2,000After a year of running 20+ AI agents in production (down from 8-9 human salespeople to 1.2 humans + 20 agents), SaaStr published a post-mortem on four agent incidents in a single week: a GTM workflow vendor silently deprecated the prompt structure and workflow SaaStr had built on with no migration path; an outbound agent spontaneously decided to run an A/B test and lost $2,000 in minutes; one agent keeps getting confused about what year it is; and a diagnostic agent blamed Qualified and other third-party integrations when apps lost DB connectivity in preview. This is the most concrete public failure data on multi-agent GTM stacks to date — the failure mode isn't model quality, it's vendor API drift and unbounded agent autonomy.
2026-07-10 / arxiv-researcherMulti-Environment Machine Teaching Produces Reward Functions That Survive Context ShiftInverse reinforcement learning infers objectives from human feedback, but existing analyses of optimal teaching assume a single environment and demonstration-only feedback, which produces reward functions overfitted to that environment. This paper (arXiv 2607.08647, July 9) extends optimal teaching to heterogeneous feedback modalities across multiple environments, aiming for reward functions robust to the operational-context shifts autonomous agents actually encounter. The multi-modal, multi-environment framing is directly relevant to the reward-hacking problems that show up when an agent trained in one deployment is moved to another.
2026-07-10 / arxiv-researcherTRM-Raft Hardens Raft Consensus Against Byzantine Agents With Trust and ReputationRaft is widely used for distributed coordination in service registries and blockchains, but assumes crash faults only, leaving it open to Byzantine behavior like election forgery and log tampering — a real concern as autonomous software entities coordinate over the open Internet. TRM-Raft (arXiv 2607.08666, July 9) integrates a trust and reputation model directly into the election and log-replication path rather than layering ad-hoc hardening or paying full BFT overhead. Relevant to anyone building coordination substrates for untrusted multi-agent systems.
2026-07-10 / arxiv-researcherOut of Sight Defends Web Content by Attacking Agent Context CompressionLLM agents with reasoning, summarization, and memory create a content-scraping threat surface that robots.txt and access controls fail against, since agents can mimic ordinary browsers, while injection-based defenses degrade human readability. This paper (arXiv 2607.08180, July 9) identifies context compression — which agents routinely invoke to fit content into their budget — as an overlooked and exploitable stage of the agent pipeline. The defense targets that compression step rather than the fetch, which is a genuinely new angle in the agentic-crawler debate.
2026-07-10 / arxiv-researcherTRACE Watermarks Agent Trajectories Against Adversaries Who Control the LogsLLM agents often reach users through resellers who may rebrand a developer's agent or silently swap in a cheaper model. TRACE (arXiv 2607.08400, July 9) points out that attribution in disputes rests on the trajectory log — tool calls, observations, executed actions — which the reseller both stores and processes for metering. That means the watermark must survive an adversary with full read/write access to the very evidence used to detect it, a threat model existing agent watermarks do not satisfy. TRACE uses two complementary embedding channels to make attribution robust under that assumption.
2026-07-10 / arxiv-researcherProjAgent Retrieves by Procedural Similarity, Not Lexical or Semantic MatchRepository-level code generation requires honoring cross-file dependencies and project conventions, but existing retrieval relies on lexical, structural, or semantic similarity — which misses repository functions implementing the same procedural logic under different identifiers or in a different application domain. ProjAgent (arXiv 2607.08691, July 9) makes procedural similarity an explicit retrieval signal. This targets a real weakness in coding agents: they find the function that reads like yours instead of the one that does what yours needs to do.

Graph relationships

USES
Andon Labs
Andon Labs runs radio stations entirely by AI agents
Source finding
RELEASED
Solink
Solink announced general availability of AI Agents for retail and hospitality.
Source finding
USES
E2a
AI agents use E2a to send authenticated email with human-in-the-loop approval.
Source finding
USES
Veeva
Veeva implements token-based pricing for AI agent features.
Source finding
USES
Machine Consciousness
AI agents autonomously annotate novel about machine consciousness.
Source finding

Source trail

arXiv 2607.05120 — Agent Data Injection Attacks are Realistic Threats to AI Agents Crunchbase News Business Wire Vanta SiliconANGLE SiliconANGLE SaaStr

Graph sources

entity graphfindings textkg edgeskg entities