"Author, run, debug, and locally test Prefect flows and tasks with
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill flow-task-authoring --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Flow Task Authoring?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-flow-task-authoring)More formats (shields.io, HTML) on the badges page.
---
name: flow-task-authoring
description: "Author, run, debug, and locally test Prefect flows and tasks with
retries, caching, states, futures, task runners, results, transactions, hooks,
and concurrency."
disable-model-invocation: true
metadata:
disco-role: operating
license: Apache 2.0
---
# Flow And Task Authoring
Use this sub-skill when a task is about writing or debugging Prefect SDK workflow code: `@flow`, `@task`, local execution, state handling, futures, task runners, result persistence, retries, caching, logging, hooks, interactive pauses, transactions, and unit-style tests.
## Start Here
1. Read [references/api-reference.md](references/api-reference.md) for live Prefect 3.6.24 decorator parameters, state constructors, futures, task runners, cache policies, result storage, pause/resume, transactions, and concurrency contexts.
2. Read [references/workflows.md](references/workflows.md) for copyable patterns: basic flows, task composition, `.submit()`, `.map()`, state inspection, local tests, cached tasks, result persistence, retry policies, hooks, and nested subflows.
3. Read [references/troubleshooting.md](references/troubleshooting.md) before diagnosing parameter validation, cache misses, retry behavior, missing logs, unresolved futures, failed states, result persistence, async/sync mismatches, pause/suspend behavior, timeouts, and concurrency-limit failures.
4. Run `python scripts/flow_task_smoke.py --help` when you need a deterministic local smoke example with no network, credentials, long-running service, or original-repo dependency.
## Routing Boundaries
- Use this sub-skill for authoring and local SDK behavior inside Python files, including `Flow` and `Task` objects created by decorators.
- Route deployment creation, `flow.serve`, `flow.deploy`, schedules, work pools, work queues, workers, and `prefect.yaml` to `../deployments-workers/SKILL.md`.
- Route CLI profiles, server startup/status, Cloud login, dashboard, variables, artifacts, and operational commands to `../cli-server-operations/SKILL.md`.
- Route direct `get_client`, `PrefectClient`, schema models, profiles/settings internals, and `temporary_settings` to `../api-client-settings/SKILL.md`.
- Route maintainer-only repository changes, test selection, `uv`/`just`, generated docs, and source-tree development workflow to `../repo-development/SKILL.md`.
## Safe Defaults
- Prefer direct local calls to `flow_fn(...)` or `task_fn(...)` while authoring; add deployments/workers only after the workflow behavior is correct.
- Prefer `@flow(log_prints=True)` or `get_run_logger()` when the user expects logs to appear in Prefect run logs.
- Prefer `.submit()` or `.map()` only inside a flow when concurrency or dependency tracking is needed; call tasks directly when sequential behavior is clearer.
- Prefer `future.result()` or `PrefectFutureList.result()` before returning or asserting values; use `return_state=True` when debugging failed states without raising immediately.
- Prefer explicit `persist_result=True`, `result_storage_key`, or `cache_policy=INPUTS` when cache/result reuse must survive beyond one task call.
## Bundled Script
- `scripts/flow_task_smoke.py` demonstrates deterministic flow/task authoring, local HTML parsing adapted from the repo example without network, retries, cache configuration, futures, mapping, hooks, and state inspection.
- The script prints JSON and supports `--mode basic`, `--mode futures`, `--mode state`, `--mode cache`, and `--mode all` for focused checks.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!