Short answer: - Those files in

~/.prefect/storage/

are persisted results and/or cached task results. - Yes, it’s safe to delete them when you don’t need to reuse results/caches and no flows are currently running. Deleting them will force recomputation on future runs. - You can avoid storing them by disabling result persistence and/or caching globally or per-flow/task. Details: What they are - Prefect 3.x writes task/flow results (and cache data) to a local result store by default at

~/.prefect/storage/

unless you’ve configured a different result storage block. - The filenames are storage keys (UUID-/hash-like strings) and hold serialized results and metadata. - They are not your source code or deployments—just result artifacts. - Docs: Results, Caching, Settings Is it safe to delete? - Yes, with caveats: - Do not delete while flows are running. - If you rely on cached results across runs, you’ll lose those caches and tasks will recompute. - If you plan to load persisted results from prior runs (e.g., inspecting state results later), deletion will make them unavailable. How to stop storing results - Globally (recommended if you don’t want local result files at all): - Keep global persistence off (it’s off by default), and disable caching. - You can enforce via Prefect config:

prefect config set PREFECT_RESULTS_PERSIST_BY_DEFAULT=false
    prefect config set PREFECT_TASKS_DISABLE_CACHING=true

- Optionally move the directory to an ephemeral path if you still want caching but don’t want long-lived files:

prefect config set PREFECT_RESULTS_LOCAL_STORAGE_PATH=/tmp/prefect-results

- Per-task/flow: - Turn off persistence and caching explicitly and/or send results to a Null sink:

from prefect import flow, task
    from prefect.cache_policies import NO_CACHE
    from prefect.filesystems import NullFileSystem

    @task(persist_result=False, cache_policy=NO_CACHE, result_storage=NullFileSystem())
    def compute():
        return "no files written"

    @flow(persist_result=False, result_storage=NullFileSystem())
    def my_flow():
        return compute()

- If you just want to disable caching but allow in-memory results during a run:

@task(cache_policy=NO_CACHE)
    def compute():
        ...

Cleaning up existing files - Safe when nothing is running:

rm -rf ~/.prefect/storage/*

- Or move them:

mv ~/.prefect/storage ~/.prefect/storage.backup

If you want, tell me whether you’d like to disable storage globally for all runs or just certain flows/tasks, and if you need caching at all—I can tailor the exact config or code changes.