Web research, research summaries, onchain analysis, classification, policy checks, human approval and memory.
Skills let a workflow research, decide and remember, with guardrails: Policy Guard blocks actions outside your rules, and Human Approval waits for your OK.
Web ResearchCloud
Search, crawl, sitemap, or API-backed web research in one node.
Inputs: Trigger, Query, URL, API Key. Outputs: Results, URLs, Pages, Ranked, Citations, Text, Raw, Errors, Count.
Research AgentCloud
Analyze gathered research into findings, citations, confidence, retry guidance, and a decision-ready report.
Inputs: Trigger, Pages, Text, Question, API Key. Outputs: Summary, Findings, Recommendation, Citations, Report, Confidence, Quality, Retry, Retry Query, Missing, Reason.
Onchain AnalystCloud
Verify and analyze Stellar accounts, balances, holders, orderbooks, liquidity pools, assets, operations, and intent into reusable market/action insight.
Inputs: Trigger, Market Data, Orderbook, Liquidity Pool, Asset, Account, Amount, Intent, API Key. Outputs: Usable, Review, Avoid, Insight, Recommendation, Report, Confidence, Risk, Reason.
Source VerifierCloud
Check research sources for trusted domains, official docs, and score thresholds.
Inputs: Trigger, Pages, Citations. Outputs: Verified, Review, Blocked, Report, Score.
Decision GateCloud
Route agent logic using confidence, risk, recommendation, and proceed/block phrases.
Inputs: Trigger, Confidence, Risk, Recommendation. Outputs: Proceed, Ask Human, Stop, Decision, Reason.
Policy GuardCloud
Block or review actions that violate spend, asset, destination, contract, or network policy.
Inputs: Trigger, Amount, Asset, Destination, Contract. Outputs: Approved, Review, Blocked, Report.
TX Simulator
Estimate fee, operation count, and risk before submitting a Stellar transaction.
Inputs: Trigger, TX Env, Operations, Destination, Asset, Amount. Outputs: Safe, Risky, Simulation, Risk, Fee.
Human Approval
Route an agent action through pending, approved, or rejected human review state.
Inputs: Trigger, Request, Approver. Outputs: Approved, Rejected, Pending, Approval.
Save MemoryCloud
Persist structured agent memory in a namespace and key.
Inputs: Trigger, Value, Key. Outputs: Done, Stored.
Recall MemoryCloud
Read structured agent memory by namespace and key.
Inputs: Trigger, Key. Outputs: Found, Missing, Value, Record.
ClassifierCloud
Classify text into categories using LLM or rule-based matching, routing to trigger outputs.
Inputs: In, Text, API Key Override. Outputs: Category: positive, Category: negative, Category: neutral, Matched Category, Confidence, Reason, Error Msg, Failed.
