ArcadeDB is a database management system that stores and works with several kinds of data, including graph, document, key-value, time-series, geospatial, and vector data. It is used by applications that need one database supporting interfaces such as SQL, Cypher, Gremlin, HTTP/JSON, MongoDB, and Redis. The catalogue add-on operates ArcadeDB.
Borrowing it
Nothing to install: this file belongs to ArcadeData/arcadedb. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ArcadeData/arcadedb/main/.claude/skills/engine-concurrency/SKILL.mdgit clone --depth 1 https://github.com/ArcadeData/arcadedbWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/arcadedata/arcadedb/engine-concurrency)<a href="https://agentmods.dev/skills/arcadedata/arcadedb/engine-concurrency"><img src="https://agentmods.dev/badge/skills/arcadedata/arcadedb/engine-concurrency/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/arcadedata/arcadedb/engine-concurrency"><img src="https://agentmods.dev/badge/skills/arcadedata/arcadedb/engine-concurrency.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Rogue Agent · line 41 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00079 | $0.03572 |
| Opus 5.5 | $0.00032 | $0.01429 |
| Sonnet 5.5 | $0.00016 | $0.00714 |
| Haiku 4.5 | $0.00008 | $0.00357 |
Grade A, and why
engine-concurrency scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Engine Concurrency and Parallelism
Core principle: ArcadeDB avoids the JDK common ForkJoinPool (ForkJoinPool.commonPool()) for engine-internal parallelism. The common pool is shared with user-supplied code (Gremlin, Polyglot, custom SQL functions, application JVM) and JDK internals (parallel GC, reference handler), so long-running engine work submitted there starves user code and JDK housekeeping. Engine code that needs parallelism submits to one of the dedicated pools below; the rule is documented at the head of com.arcadedb.query.QueryEngineManager's class Javadoc.
Existing JDK-common-pool callers (tagged in source with NOTE (concurrency) comments referencing the rule, will migrate as workloads justify it):
GraphBatch.parallelSort(engine/...graph/GraphBatch.java)
ArcadeStateMachine.notifyInstallSnapshotFromLeader was on this list until issue #6202 gave it a dedicated
single-worker executor (arcadedb-raft-snapshot-install); a follower snapshot install is a full database
download, i.e. the longest-running thing the HA layer does, and it no longer parks a common-pool worker for it.
Dedicated thread pools
| Pool | Module | Purpose | Sizing | Saturation policy |
|---|---|---|---|---|
QueryEngineManager JVM-wide pool |
engine |
Query-time parallelism: graph algorithms (parallelForRange), parallel index scans, anything that forks query work |
arcadedb.queryParallelismPoolThreads (default max(2, CPU)) |
Bounded queue (arcadedb.queryParallelismQueueSize, default 1024), caller-runs rejection, throttled WARNING (60s window) |
SparseVectorScoringPool |
engine |
Reserved for per-segment parallel scoring of LSM_SPARSE_VECTOR top-K (dispatch wiring deferred to issue #4085) |
arcadedb.sparseVectorScoringPoolThreads (default max(2, CPU)), 1024 queue; lazy-init via Holder idiom |
Bounded queue, caller-runs, throttled WARNING |
ParallelScanProducerPool |
engine |
The BLOCKING producer tasks of a parallel bucket scan (FetchFromTypeExecutionStep.syncPullParallel) |
arcadedb.parallelScanProducerPoolThreads (default max(2, CPU)) |
Deviation: unbounded queue and NO caller-runs - caller-runs on a blocking producer is the #4948 self-deadlock. Back-pressure comes from each query's bounded RESULT queue; queue_depth is the saturation signal |
AsyncCommandPool |
engine |
Commands dispatched with awaitResponse=false that parse to DDL, which cannot run on an async worker (#6303 item 3) |
arcadedb.asyncCommandPoolThreads (default max(2, CPU)), arcadedb.asyncCommandQueueSize (default 1024) |
Bounded queue, caller-runs even on a shut-down pool (there is no future to cancel and the submitter has already counted the command as in flight), throttled WARNING |
DatabaseAsyncExecutor |
engine |
Per-database async ops (background scheduled tasks, async commit) | arcadedb.asyncWorkerThreads (default CPU - 1, min 1) |
Bounded queue (arcadedb.asyncOperationsQueueSize, default 1024) |
PageManagerFlushThread |
engine |
Dedicated single thread for paginated-component page writes | 1 thread | Backpressure via arcadedb.maxRAMForPageRamUsageInMB |
TransactionManager.HousekeepingScheduler |
engine |
Periodic WAL housekeeping (rotation, idle WAL release, drop of retired WAL files) of EVERY database of the JVM, one fixed-delay task per database (issue #10234; it used to be a java.util.Timer thread per database) |
2 daemon threads ArcadeDB WAL Housekeeping-N, JVM-wide, lazy (Holder idiom) |
n/a: a fixed-delay task per open database, removed on cancel; a slow pass delays only the next pass of the same database while the other thread serves the rest. Never shut down (daemon threads) and not wired to Micrometer: a pass over 10 s logs a WARNING, over one period a FINE line; measured at about 1 us per database per tick |
TimeSeriesEngine pool |
engine |
Time-series rollup work | configurable | n/a |
BackupScheduler |
server |
Cron-style backup jobs | scheduled executor | n/a |
MaterializedViewScheduler |
server |
Cron-style MV refresh | scheduled executor | n/a |
| Raft HA pools | ha-raft |
Leader election, log replication | configurable in Raft conf | n/a |
ArcadeStateMachine.snapshotInstallExecutor |
ha-raft |
Leader-initiated follower snapshot install, off the Ratis state-machine thread | 1 worker, 16-deep queue (Ratis serialises installs per division) | Abort, turned into a failed future so Ratis retries - never caller-runs, which would put the download back on the Ratis thread |
IndexWarmUpScheduler |
engine |
Warms, in the background, the vector indexes a schema reload re-created cold (issue #10021): a debounced daemon thread that exists only while there is work, one index at a time | 1 thread per LocalSchema, arcadedb.vectorIndex.warmUpAfterSchemaReloadDelayMs debounce (default 1 s, -1 off) |
Deviation: not a DedicatedThreadPool, whose caller-runs saturation would run the load on the Ratis apply thread this work is moved off. A newer request replaces a waiting one and abandons the rest of a running batch, so nothing queues; close() interrupts and joins for at most 5 s |
ServerSecurity.permissionsRefreshExecutor |
server |
Re-deriving the cached per-database permissions after a group document arrives over HA replication, off the Raft state-machine apply thread (issue #7510) | core 0 (an idle server carries no thread) / max 1, 30 s keep-alive, 1-deep queue | Drop, logged at FINE - and that is coalescing, not loss: the task reads the group document when it RUNS, and the new document is published before the submit, so the one already queued covers every change dropped behind it |
QueryWarmUp |
server |
Warms the query stack once the server is ONLINE - a SQL parse, SELECT 1 and RETURN 1 on an open database - so the first client request does not pay for loading it (issue #10218) |
1 daemon thread ArcadeDB Query Warm-Up at MIN_PRIORITY, started once and ended when done; arcadedb.server.queryWarmUp (default on) |
n/a: runs once |
| Undertow IO + worker | server |
HTTP request handling | hardcoded 500 worker threads | Undertow built-in |
ServerMonitor |
server |
Periodic metric collection | scheduled executor | n/a |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago Changed · +1 lines e1ec3adca9c8
- 14d ago Changed · +1 lines b1ff61054681
- 1mo ago Changed · +6 lines e3e05f0c4bf2
- 1mo ago Changed · +1 lines 94762e09e4f3
- 2mo ago First seen · 117 lines · 79 tokens per session scan A 8ab88048bda2
engine-concurrency is a skill published in the GitHub repository ArcadeData/arcadedb (1,225 stars, last pushed today), licensed Apache-2.0. It adds 79 tokens to every session and 3,572 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
arangodb
ArangoDB multi-model database reference covering document, graph, and key-value operations. Includes AQL query language, graph traversals, shortest path algorithms, indexing, ArangoSearch, clustering, backup, and administration.
data-model-selector
Choose between relational, document, and graph data models for an application by analyzing data shape, relationship complexity, and query patterns. Use when asked "should I use MongoDB or PostgreSQL?", "when does a graph database make sense?", "how do I choose between SQL and NoSQL?", or "what data model fits my…
triage-issues
Triage GitHub issues in the googleapis/mcp-toolbox repo: propose the correct labels (type / priority / product / status), check for duplicates, verify a bug has enough info to act on, and draft a triage comment. Use whenever a maintainer asks you to triage, label, categorize, prioritize, or "look at" an issue (or a…
postgresql-indexing
PostgreSQL indexing best practices for Prowler: index design, partial indexes, partitioned table indexing, EXPLAIN ANALYZE validation, concurrent operations, monitoring, and maintenance. Trigger: When creating or modifying PostgreSQL indexes, analyzing query performance with EXPLAIN, debugging slow queries, reviewing…
alert-rule-troubleshoot
This skill should be used when the user reports that an alert rule is "not firing", "no alert was sent", "the rule didn't trigger", "the rule isn't working", "it should have alerted but didn't", "why didn't I get an alert", "alert rule not firing", or wants to diagnose why a specific alert rule failed to produce an…
ops-troubleshooting
This skill should be used when the user asks to "troubleshoot", "diagnose", "debug alert", "investigate incident", "locate a fault", "investigate an alert", "diagnose a problem", "fix an issue", "check alerts", "analyze alerts", "root cause analysis", "check metrics", "check logs", or discusses…