Agent Loop Detector

agent.loop_detect

ProductionBounded custom (Class B)Deterministicstrictfastv1.0.0

No network

Overview

You send 2 to 5,000 events with action, optional target, stateFingerprint and outcome. window (2-200, default 20) and repeatThreshold (2-20, default 3) bound the search. similarity EXACT uses all four fields; ACTION_TARGET ignores state and outcome. We pick the earliest span, then the shortest cycle. We do not stop an agent or judge task success.

An agent that retries the same click against an unchanged page burns quota without a typed loop signal.

Reliability

Deterministic earliest-then-shortest selection; no clock or network. LOOP_EVENTS_INVALID — trigger: fewer than 2 events or an unknown similarity; retryable: no; next: send two or more closed events. LOOP_EVENT_LIMIT_EXCEEDED — trigger: more than 5,000 events; retryable: no; next: window the trace.

Context footprint

Span indices and cycle length only. A successful call costs 1 execution credit.

When to use it

When a trace of actions might be cycling and you want the earliest shortest span.

When you need to halt the agent, compare embeddings, or explain why the task failed.

Scope

Note

Detects repeated action/state spans in a bounded agent event list using EXACT or ACTION_TARGET similarity.

Non-goals

Does not stop an agent
Does not use embeddings
Does not judge task success

Properties

IdempotentRetry safeStreaming supportedFile input supportedRequires external service

Capabilities

analysis

Cost & billing

Flat rate per invocation

MetercallsBillable onsuccessCost per call1 credit

File handling

Accepts file_refProduces file_refMax input2 MB

How we handle your data

Sensitive input, safe output

What this means

This skill may receive sensitive content, but it does not target sensitive data and returns only a structured non-sensitive result.

Squillify guarantees

Runs inside Squillify
This skill runs inside Squillify.
Payloads are not sent to an external service by this skill.
Squillify does not permanently store submitted payloads.
Telemetry records metadata only, not submitted content.
Submitted payloads are not used for training.
Submitted payloads are not reviewed by humans.
This skill does not target sensitive data.
The response does not contain submitted content.

Your responsibility

Squillify does not determine whether submitted content is sensitive. Your application is responsible for deciding what data it is allowed to submit. Your application is responsible for securing the response.

Read the Microskill Data Handling Policy

Input schema

Field	Type	Required	Description
events	array	Yes	Bounded array of 1 through 10001 events; runtime requires 2 and accepts at most 5000.
window	integer	No	Cycle length ceiling from 2 through 200. Omit it for 20.
repeatThreshold	integer	No	Consecutive copies required, from 2 through 20. Omit it for 3.
similarity	string	No	EXACT or ACTION_TARGET. Omit it for EXACT.

Output schema

Field	Type	Required	Description
loopDetected	boolean	Yes	Whether a span met the threshold.
similarity	string	Yes	Applied similarity rule EXACT or ACTION_TARGET after defaulting.
window	integer	Yes	Applied cycle-length ceiling after defaulting omitted values.
repeatThreshold	integer	Yes	Applied consecutive-copy threshold after defaulting omitted values.
cycleLength	integer	Yes	Chosen cycle length, or 0.
confidence	string	Yes	NONE, MEDIUM or HIGH for the chosen span.
eventIndices	array	Yes	Bounded array of 0 through 5000 indices covering the chosen span.
spans	array	Yes	Bounded array of 0 through 8 chosen spans; v1 reports at most one.

API contract

POST /v1/execute

curl

curl -sS -X POST "$BASE/v1/execute" -H "Authorization: Bearer $KEY" -H 'Content-Type: application/json' -d '{"skillKey":"agent.loop_detect","input":{"events":[{"action":"a"},{"action":"a"},{"action":"a"}]}}'

Integrations

TypeScript SDK

import { Squillify } from "@squillify/sdk";

const sq = new Squillify({ apiKey: process.env.SQUILLIFY_API_KEY });

const { output } = await sq.execute("agent.loop_detect", {"events":[{"action":"search","target":"web","outcome":"ok","stateFingerprint":"s1"},{"action":"click","target":"r1","outcome":"ok","stateFingerprint":"s2"},{"action":"search","target":"web","outcome":"ok","stateFingerprint":"s1"},{"action":"click","target":"r1","outcome":"ok","stateFingerprint":"s2"},{"action":"search","target":"web","outcome":"ok","stateFingerprint":"s1"},{"action":"click","target":"r1","outcome":"ok","stateFingerprint":"s2"}]});

Errors

Code	HTTP	Retryable	Description
VALIDATION_ERROR	400	No	The request payload failed schema validation.
UNAUTHORIZED	401	No	Missing or invalid API key.
SKILL_NOT_ACCESSIBLE	403	No	The API key is not granted access to this skill.
SKILL_NOT_AVAILABLE_FOR_PLAN	403	No	This skill is not included in your current plan.
PLAN_ELIGIBILITY_UNAVAILABLE	503	Yes	Plan eligibility could not be decided. Retry after backoff.
SKILL_NOT_FOUND	404	No	The requested skill or version does not exist.
TENANT_RATE_LIMITED	429	Yes	Request rate limit exceeded. Retry after backoff.
QUOTA_EXCEEDED	429	Yes	Plan quota exhausted. Retryable once the quota period resets.
EXECUTION_TIMEOUT	504	Yes	Execution exceeded the allowed time budget.
LOOP_EVENTS_INVALID	400	No	The request payload failed this skill's input validation.
LOOP_EVENT_LIMIT_EXCEEDED	429	No	A size, quota, rate, or time limit was exceeded.

Other limits

Limit	Value
Recommended inline bytes	32 KiB

Limitations

Does not stop an agent
Does not use embeddings
Does not judge task success
Halting a running agent
Semantic similarity of free text

How to call

POST /v1/execute (requires authentication)

Request bodyjson

{
  "skillKey": "agent.loop_detect",
  "input": {
    "events": [
      {
        "action": "search",
        "target": "web",
        "outcome": "ok",
        "stateFingerprint": "s1"
      },
      {
        "action": "click",
        "target": "r1",
        "outcome": "ok",
        "stateFingerprint": "s2"
      },
      {
        "action": "search",
        "target": "web",
        "outcome": "ok",
        "stateFingerprint": "s1"
      },
      {
        "action": "click",
        "target": "r1",
        "outcome": "ok",
        "stateFingerprint": "s2"
      },
      {
        "action": "search",
        "target": "web",
        "outcome": "ok",
        "stateFingerprint": "s1"
      },
      {
        "action": "click",
        "target": "r1",
        "outcome": "ok",
        "stateFingerprint": "s2"
      }
    ]
  }
}

Examples

A search-click cycle repeating three times

Inputjson

{
  "events": [
    {
      "action": "search",
      "target": "web",
      "outcome": "ok",
      "stateFingerprint": "s1"
    },
    {
      "action": "click",
      "target": "r1",
      "outcome": "ok",
      "stateFingerprint": "s2"
    },
    {
      "action": "search",
      "target": "web",
      "outcome": "ok",
      "stateFingerprint": "s1"
    },
    {
      "action": "click",
      "target": "r1",
      "outcome": "ok",
      "stateFingerprint": "s2"
    },
    {
      "action": "search",
      "target": "web",
      "outcome": "ok",
      "stateFingerprint": "s1"
    },
    {
      "action": "click",
      "target": "r1",
      "outcome": "ok",
      "stateFingerprint": "s2"
    }
  ]
}

Expected outputjson

{
  "spans": [
    {
      "start": 0,
      "end": 5,
      "cycleLength": 2,
      "repeats": 3
    }
  ],
  "confidence": "MEDIUM",
  "cycleLength": 2,
  "eventIndices": [
    0,
    1,
    2,
    3,
    4,
    5
  ],
  "window": 20,
  "repeatThreshold": 3,
  "similarity": "EXACT",
  "loopDetected": true
}

Tip

The earliest span of length 2 meets the default threshold of 3.

Same action with changing state is not an EXACT loop