Replace the current mix of:

• fg/bg ordering barriers
• numeric filename ordering used as dependency encoding
• marker files like cdp_url.txt, target_id.txt, extensions.json, navigation.json, prenav.json
• sibling file/log polling

with a smaller model:

• hooks subscribe only via on_<EventName>__...
• hooks emit real facts on stdout as JSONL
• abx-dl routes those facts dynamically through abxbus
• abx-dl auto-emits After<EventName> as a synthetic settle barrier
• abx-dl reduces all prior scoped events into a derived key/value context
• abx-dl mirrors crawl-scoped context to CRAWL_DIR/derived.env
• abx-dl mirrors snapshot-scoped context to SNAP_DIR/derived.env by overlaying snapshot state on top of the current crawl context
• hooks receive the full reduced context as env vars
• hooks receive only the triggering event payload as CLI args

Chrome remains the stable namespace for all Chrome-like providers.

Use only:

• on_<EventName>__...

Examples:

• on_SnapshotChromeTabReady__21_consolelog.daemon.bg.js
• on_AfterSnapshotChromeTabReady__30_chrome_navigate.js
• on_SnapshotChromeTabNavigated__51_screenshot.js
• on_AfterSnapshot__70_parse_html_urls.py

There is no separate after_* hook syntax.

Emitted by hooks when they create new meaning or new payload.

Examples:

• CrawlChromeBrowserReady
• CrawlChromeExtensionsReady
• SnapshotChromeBrowserReady
• SnapshotChromeExtensionsReady
• SnapshotChromeTabReady
• SnapshotChromeTabNavigated
• SnapshotChromeTabNavigationFailed
• SnapshotChromeMainResponseSaved
• SnapshotChromeStaticFileHandled
• SnapshotPdfOcrComplete
• UrlDiscovered

Emitted only by abx-dl.

Examples:

• AfterSnapshot
• AfterSnapshotChromeTabReady
• AfterSnapshotChromeTabNavigated
• AfterSnapshotPdfOcrComplete

Meaning:

• After<Event> = the full subtree rooted at <Event> has settled

After<Event> is for ordering only. It does not introduce new semantics.

Use a real event when:

• a hook creates a meaningful state transition
• a hook creates new payload another hook may consume
• a hook wants to update the shared derived context

Use After<Event> when:

• a later hook only needs barrier/ordering semantics
• no new domain payload is needed

Examples:

• chrome_navigate should run on AfterSnapshotChromeTabReady once pre-navigation bg settle semantics are defined precisely
• screenshot should run on SnapshotChromeTabNavigated, not AfterSnapshotChromeTabReady
• late parsers/indexers should run on AfterSnapshot
• OCR should emit a real event like SnapshotPdfOcrComplete
• late OCR consumers can run on AfterSnapshotPdfOcrComplete

abx-dl must not know Chrome-specific meanings.

It only needs to know:

• how to route events by string name
• how to launch hooks for on_<Event>
• how to synthesize After<Event>
• how to route both real events and synthetic After<Event> barriers through the same generic dispatch path
• how to reduce scoped prior events into current key/value context
• how to mirror reduced context into derived.env
• how to pass env vars and event payload args into hooks
• how to wait for the full root Snapshot tree before cleanup

No plugin-specific event classes or Chrome-specific scheduling logic should be added to abx-dl.

Hooks emit JSONL records with:

• required top-level type
• optional normal event payload fields
• optional reserved env patch for durable aggregate state

Example:

{
  "type": "SnapshotChromeTabReady",
  "url": "https://example.com",
  "env": {
    "CDP_URL": "ws://127.0.0.1:9222/devtools/browser/...",
    "TARGET_ID": "ABC123",
    "EXTENSIONS": [
      {"id": "ublock", "name": "uBlock Origin"}
    ]
  }
}

env is the only generic state-update lane.

Rules:

• scalar values: last write wins
• null: unset/remove key
• arrays: append
• objects: deep-merge
• type changes: latest value replaces incompatible earlier type

Implications:

• multiple hooks can contribute to EXTENSIONS
• later hooks can correct stale TARGET_ID / CDP_URL
• later hooks launched by abx-dl will automatically see corrected values

For derived.env mirroring:

• scalars are written as plain env values
• arrays/objects are serialized as JSON strings

Examples:

• CDP_URL=ws://...
• TARGET_ID=ABC123
• EXTENSIONS=[{"id":"ublock","name":"uBlock Origin"}]

Only abx-dl writes derived.env.

The event log is canonical.

abx-dl should maintain scope directionality explicitly:

• crawl events reduce into one crawl context mirrored to CRAWL_DIR/derived.env
• each snapshot starts from the current crawl context
• snapshot events then overlay snapshot-local state on top of that base
• the resulting per-snapshot view is mirrored to SNAP_DIR/derived.env
• snapshot-local reduction must never write back into CRAWL_DIR/derived.env

derived.env is:

• generated
• for observability and crash inspection
• useful for manual standalone hook invocation
• not the canonical source of truth

This one-way flow is required because multiple snapshots may run in parallel within a single crawl.

Only durable/current-state values that later hooks may need fresh:

• URL
• CRAWL_ID
• SNAPSHOT_ID
• CRAWL_DIR
• SNAP_DIR
• CDP_URL
• TARGET_ID
• FINAL_URL
• HTTP_STATUS
• EXTENSIONS
• MAIN_RESPONSE_PATH
• MAIN_RESPONSE_MIME
• STATICFILE_HANDLED
• STATICFILE_PATH

Avoid stuffing transient/debug-only values into reduced context.

Every hook gets the full reduced context as env vars.

Examples:

• URL
• CRAWL_ID
• SNAPSHOT_ID
• SNAP_DIR
• CDP_URL
• TARGET_ID
• FINAL_URL
• HTTP_STATUS
• EXTENSIONS

Every hook also gets the triggering event payload as --kebab-case=value CLI args.

Important split:

• reduced context is passed via env vars
• current event payload is passed via CLI args
• only fields on the triggering event become CLI args
• reduced context values should not also be sprayed onto the CLI unless they are part of the triggering event payload

Example:

• env:
  • URL=...
  • CDP_URL=...
  • TARGET_ID=...
• CLI:
  • --url=...
  • --final-url=...
  • --http-status=200

EXTRA_CONTEXT stays opaque pass-through metadata only.

• hooks should not depend on it for runtime behavior
• hooks should consume explicit env vars / CLI args instead
• emitted ArchiveResult, Snapshot, and UrlDiscovered records may still merge it forward for lineage

• make root lifecycle event types line up with hook filenames:
  • CrawlSetup
  • Snapshot

• keep parse_hook_filename() as the source of truth for on_<Event>__...
• no whitelist of supported event names
• store the raw trigger event string exactly as parsed from the filename

• keep the existing service
• dispatch all crawl-tree hooks dynamically by event string, including synthetic AfterCrawl* events if/when used
• convert hook stdout records with type beginning with Crawl into BaseEvent(event_type=...)
• reduce prior crawl-scoped events into current crawl context
• mirror reduced crawl context to CRAWL_DIR/derived.env
• pass reduced crawl context as env vars plus current event payload as CLI args
• synthesize After<Event> for crawl events

• keep the existing service
• dispatch all snapshot-tree hooks dynamically by event string, including synthetic AfterSnapshot* events
• convert hook stdout records with type beginning with Snapshot into BaseEvent(event_type=...)
• route non-reserved emitted events like UrlDiscovered as normal bus events under the current snapshot tree instead of hardcoding exact type == "Snapshot"
• start from the current crawl context as a base
• reduce prior snapshot-scoped events into a snapshot-local overlay on top of that base
• mirror reduced snapshot context to SNAP_DIR/derived.env
• pass reduced snapshot context as env vars plus current event payload as CLI args
• auto-emit After<Event> after the routed event subtree settles
• emit SnapshotCleanupEvent only after the full root Snapshot tree settles

• expand fallback ArchiveResult synthesis beyond hooks whose names start with on_Snapshot
• on_AfterSnapshot__... hooks must also get the same fallback result behavior
• more generally, fallback result synthesis should apply to hooks whose process is a descendant of a root snapshot lifecycle event, not only to one filename prefix

Rules:

1. After<Event> is emitted only by abx-dl
2. hooks must never emit After* themselves
3. After<Event> is emitted as a child of <Event>, before <Event> is considered complete
4. do not auto-generate AfterAfter<Event>
5. SnapshotCleanup waits for the full root Snapshot tree, including all nested After<Event> branches
6. AfterSnapshot is solid because it is rooted in full snapshot-tree settlement
7. AfterSnapshotChromeTabReady must not be treated as implemented until bg pre-navigation hook settle semantics are defined precisely; otherwise it can fire either too early or too late under the current bg process model

That is what makes multi-stage settled pipelines work.

• CrawlSetup
• Snapshot

• UrlDiscovered

UrlDiscovered should replace hooks emitting new Snapshot records for crawl discovery.

Payload should carry all useful metadata in one record, e.g.:

• url
• title
• tags
• bookmarked_at
• parser/feed/bookmark metadata

• CrawlChromeBrowserReady
• CrawlChromeExtensionsReady

• SnapshotChromeBrowserReady
• SnapshotChromeExtensionsReady
• SnapshotChromeTabReady
• SnapshotChromeTabNavigated
• SnapshotChromeTabNavigationFailed
• SnapshotChromeTabClosed

Only add these if they materially simplify orchestration:

• SnapshotChromeMainResponseSaved
• SnapshotChromeStaticFileHandled
• SnapshotChromeHeadersCaptured
• SnapshotChromeRedirectsCaptured

Do not add a trivial event for every implementation detail.

• only emit CrawlChrome* facts during crawl setup
• only emit SnapshotChrome* facts during snapshot execution
• never synthesize snapshot aliases from crawl browser facts

If CHROME_ISOLATION=crawl:

• crawl launch emits CrawlChromeBrowserReady and CrawlChromeExtensionsReady
• snapshot phase starts at SnapshotChromeTabReady
• there is no SnapshotChromeBrowserReady alias

If CHROME_ISOLATION=snapshot:

• snapshot launch emits SnapshotChromeBrowserReady and SnapshotChromeExtensionsReady
• snapshot phase then proceeds to SnapshotChromeTabReady

The first snapshot-scoped Chrome event guaranteed in both modes is:

• SnapshotChromeTabReady

1. CrawlSetup
2. on_CrawlSetup__89_chrome_kill_zombies.js
3. on_CrawlSetup__90_chrome_launch.daemon.bg.js in crawl isolation
   • emits CrawlChromeBrowserReady
   • emits CrawlChromeExtensionsReady
   • patches env with browser-level values like CDP_URL, EXTENSIONS
4. crawl-scoped extension config hooks subscribe to CrawlChromeExtensionsReady
   • abx_plugins/plugins/twocaptcha/on_CrawlSetup__95_twocaptcha_config.js
   • abx_plugins/plugins/claudechrome/on_CrawlSetup__96_claudechrome_config.js

1. Snapshot
2. root non-Chrome downloaders stay on Snapshot, e.g.:
   • abx_plugins/plugins/ytdlp/on_Snapshot__02_ytdlp.finite.bg.py
   • abx_plugins/plugins/gallerydl/on_Snapshot__03_gallerydl.finite.bg.py
   • abx_plugins/plugins/forumdl/on_Snapshot__04_forumdl.finite.bg.py
   • abx_plugins/plugins/git/on_Snapshot__05_git.finite.bg.py
   • abx_plugins/plugins/wget/on_Snapshot__06_wget.finite.bg.py
   • abx_plugins/plugins/archivedotorg/on_Snapshot__08_archivedotorg.finite.bg.py
   • abx_plugins/plugins/favicon/on_Snapshot__11_favicon.finite.bg.py
   • abx_plugins/plugins/papersdl/on_Snapshot__66_papersdl.finite.bg.py
3. Chrome branch:
   • crawl isolation: chrome_tab subscribes directly to Snapshot
   • snapshot isolation: launch emits SnapshotChromeBrowserReady and SnapshotChromeExtensionsReady, then chrome_tab runs
4. chrome_tab emits SnapshotChromeTabReady
   • patches env with TARGET_ID and any corrected tab-level state
5. pre-navigation listeners run on SnapshotChromeTabReady
6. once the pre-navigation bg settle semantics are defined correctly, chrome_navigate can run on AfterSnapshotChromeTabReady
7. chrome_navigate emits SnapshotChromeTabNavigated or SnapshotChromeTabNavigationFailed
   • patches env with corrected FINAL_URL, HTTP_STATUS, TARGET_ID if needed
8. post-navigation extractors run on SnapshotChromeTabNavigated
9. once the entire root Snapshot tree settles, abx-dl emits AfterSnapshot
10. late parsers / OCR / indexers / cleanup / hashes run on AfterSnapshot
11. if any of those late hooks emit new real events, abx-dl may synthesize After<ThoseEvents> and the tree continues
12. SnapshotCleanup runs only after the full root Snapshot tree finishes

AfterSnapshot solves these generically:

• Chrome-disabled runs still work
• URL parsers can run after all snapshot outputs exist
• later stages like OCR can emit real follow-on events without special scheduler code

Delete entirely:

• abx_plugins/plugins/chrome/on_CrawlSetup__91_chrome_wait.js
• abx_plugins/plugins/chrome/on_Snapshot__11_chrome_wait.js

• emit CrawlChromeBrowserReady
• emit CrawlChromeExtensionsReady
• patch env with browser-level state

• only run in snapshot isolation
• emit SnapshotChromeBrowserReady
• emit SnapshotChromeExtensionsReady
• patch env with browser-level state
• remove current crawl-isolation no-op behavior

• in crawl isolation, subscribe to Snapshot
• in snapshot isolation, subscribe to the relevant snapshot browser-ready event
• emit SnapshotChromeTabReady
• patch env with fresh tab state
• stop being the publisher of runtime marker files used for orchestration
• split or rewrite helper usage so fresh reduced env state can replace the current session-dir marker contract over time

• move to on_AfterSnapshotChromeTabReady__30_chrome_navigate.js once that settle barrier is well-defined for bg pre-navigation hooks
• remove prenav.json polling and marker-based gating
• emit SnapshotChromeTabNavigated or SnapshotChromeTabNavigationFailed
• patch env with latest navigation state

Move to on_SnapshotChromeTabReady__...:

• abx_plugins/plugins/consolelog/on_Snapshot__21_consolelog.daemon.bg.js
• abx_plugins/plugins/dns/on_Snapshot__22_dns.daemon.bg.js
• abx_plugins/plugins/sslcerts/on_Snapshot__23_sslcerts.daemon.bg.js
• abx_plugins/plugins/responses/on_Snapshot__24_responses.daemon.bg.js
• abx_plugins/plugins/redirects/on_Snapshot__25_redirects.daemon.bg.js
• abx_plugins/plugins/staticfile/on_Snapshot__26_staticfile.daemon.bg.js
• abx_plugins/plugins/headers/on_Snapshot__27_headers.daemon.bg.js
• abx_plugins/plugins/ublock/on_Snapshot__12_ublock.daemon.bg.js
• abx_plugins/plugins/istilldontcareaboutcookies/on_Snapshot__13_istilldontcareaboutcookies.daemon.bg.js
• abx_plugins/plugins/twocaptcha/on_Snapshot__14_twocaptcha.daemon.bg.js
• abx_plugins/plugins/modalcloser/on_Snapshot__15_modalcloser.daemon.bg.js

Remove:

• readiness file creation
• numeric “must run before navigate” comments
• waitForNavigationComplete(...) used as control-plane gating

Important note:

• this stage is still the main unresolved runtime detail in the plan
• background pre-navigation hooks currently launch without a precise runtime notion of “settled enough for AfterSnapshotChromeTabReady”
• do not move chrome_navigate until that settle barrier is defined correctly

Move to on_SnapshotChromeTabNavigated__...:

• abx_plugins/plugins/seo/on_Snapshot__38_seo.js
• abx_plugins/plugins/accessibility/on_Snapshot__39_accessibility.js
• abx_plugins/plugins/infiniscroll/on_Snapshot__45_infiniscroll.js
• abx_plugins/plugins/claudechrome/on_Snapshot__47_claudechrome.js
• abx_plugins/plugins/singlefile/on_Snapshot__50_singlefile.py
• abx_plugins/plugins/singlefile/singlefile_extension_save.js
• abx_plugins/plugins/screenshot/on_Snapshot__51_screenshot.js
• abx_plugins/plugins/pdf/on_Snapshot__52_pdf.js
• abx_plugins/plugins/dom/on_Snapshot__53_dom.js
• abx_plugins/plugins/title/on_Snapshot__54_title.js
• abx_plugins/plugins/parse_dom_outlinks/on_Snapshot__75_parse_dom_outlinks.js

Move to on_AfterSnapshot__... if they should run after all snapshot outputs exist and are actually intended to consume generated snapshot artifacts rather than the original input URL directly:

• abx_plugins/plugins/readability/on_Snapshot__56_readability.py
• abx_plugins/plugins/mercury/on_Snapshot__57_mercury.py
• abx_plugins/plugins/defuddle/on_Snapshot__57_defuddle.py
• abx_plugins/plugins/htmltotext/on_Snapshot__58_htmltotext.py
• abx_plugins/plugins/claudecodeextract/on_Snapshot__58_claudecodeextract.py
• abx_plugins/plugins/trafilatura/on_Snapshot__59_trafilatura.py
• abx_plugins/plugins/opendataloader/on_Snapshot__60_opendataloader.py
• abx_plugins/plugins/liteparse/on_Snapshot__61_liteparse.py
• abx_plugins/plugins/parse_html_urls/on_Snapshot__70_parse_html_urls.py
• abx_plugins/plugins/search_backend_sqlite/on_Snapshot__90_index_sqlite.py
• abx_plugins/plugins/search_backend_sonic/on_Snapshot__91_index_sonic.py
• abx_plugins/plugins/claudecodecleanup/on_Snapshot__92_claudecodecleanup.py
• abx_plugins/plugins/hashes/on_Snapshot__93_hashes.py

Review needed for:

• abx_plugins/plugins/parse_txt_urls/on_Snapshot__71_parse_txt_urls.py
• abx_plugins/plugins/parse_rss_urls/on_Snapshot__72_parse_rss_urls.py
• abx_plugins/plugins/parse_netscape_urls/on_Snapshot__73_parse_netscape_urls.py
• abx_plugins/plugins/parse_jsonl_urls/on_Snapshot__74_parse_jsonl_urls.py

These currently parse their direct input source, not sibling generated outputs, so moving them to AfterSnapshot is not just a scheduling change. They must either:

• stay on Snapshot as direct-source parsers

or:

• be rewritten to scan generated snapshot outputs first, then moved to AfterSnapshot

• abx_plugins/plugins/twocaptcha/on_CrawlSetup__95_twocaptcha_config.js
• abx_plugins/plugins/claudechrome/on_CrawlSetup__96_claudechrome_config.js

Use:

• CrawlChromeExtensionsReady in crawl isolation
• SnapshotChromeExtensionsReady in snapshot isolation via thin wrappers if needed

Do not use .configured marker files as orchestration state.

• abx_plugins/plugins/redirects/prenav.json
• CRAWL_DIR/chrome/.twocaptcha_configured
• CRAWL_DIR/chrome/.claudechrome_configured
• SNAP_DIR/chrome/url.txt

These may remain as optional debug/manual-run artifacts, but must not be required for orchestration:

• CRAWL_DIR/chrome/cdp_url.txt
• CRAWL_DIR/chrome/chrome.pid
• CRAWL_DIR/chrome/extensions.json
• SNAP_DIR/chrome/cdp_url.txt
• SNAP_DIR/chrome/target_id.txt
• SNAP_DIR/chrome/navigation.json
• SNAP_DIR/chrome/extensions.json

Replace with event payload fields plus reduced env state on:

• CrawlChromeBrowserReady
• CrawlChromeExtensionsReady
• SnapshotChromeBrowserReady
• SnapshotChromeExtensionsReady
• SnapshotChromeTabReady
• SnapshotChromeTabNavigated
• SnapshotChromeTabNavigationFailed

• abx_plugins/plugins/consolelog/console.jsonl
• abx_plugins/plugins/headers/headers.json
• abx_plugins/plugins/responses/index.jsonl
• abx_plugins/plugins/dns/dns.jsonl

These remain useful outputs, but file existence must not mean “ready”.

Remove uses of sibling stdout.log scraping such as hasStaticFileOutput(...).

Replace with a real fact like:

• SnapshotChromeStaticFileHandled

and/or reduced state keys like:

• STATICFILE_HANDLED
• STATICFILE_PATH

Add one small generic helper on both sides:

• emit_event_record("EventName", {...})

That helper should support:

• normal payload fields
• optional reserved env patch

Shrink or split helpers that currently encode file-based orchestration:

• waitForChromeSessionState(...)
• connectToPage(...)
• waitForNavigationComplete(...)

Helpers should remain useful for direct/manual invocation, but they should stop being the control plane.

abx_plugins/plugins/staticfile/on_Snapshot__26_staticfile.daemon.bg.js should stop polling responses/index.jsonl.

This is likely a deeper refactor than a simple resubscribe:

• today staticfile starts in the pre-navigation stage but later waits for responses output
• in the new model it should either be split into a pre-navigation detector and a later saver, or otherwise rewritten so it consumes a real later event cleanly

If responses needs to fan out to staticfile, use a real event such as:

• SnapshotChromeMainResponseSaved

• abx_plugins/plugins/singlefile/tests/test_singlefile.py
• abx_plugins/plugins/claudechrome/tests/test_claudechrome.py
• abx_plugins/plugins/claudecodecleanup/tests/test_claudecodecleanup.py

• abx_plugins/plugins/redirects/tests/test_redirects.py
• abx_plugins/plugins/headers/tests/test_headers.py
• abx_plugins/plugins/consolelog/tests/test_consolelog.py
• abx_plugins/plugins/staticfile/tests/test_staticfile.py
• abx_plugins/plugins/chrome/tests/chrome_test_helpers.py
• abx_plugins/plugins/chrome/tests/test_chrome.py

Prefer asserting:

• emitted events
• reduced derived.env state where useful
• resulting outputs
• final user-visible behavior

• abx_plugins/plugins/chrome/README.md

Remove “authoritative marker file” language for:

• cdp_url.txt
• target_id.txt
• navigation.json

Document:

• event-driven control plane
• synthetic After<Event> settle barriers
• reduced context mirrored to derived.env
• files as outputs/debug aids only

1. Update abx-dl for dynamic Crawl* / Snapshot* dispatch and synthetic After<Event> emission.
2. Add scoped context reduction and derived.env mirroring.
3. Pass full reduced context via env vars and current event payload via CLI args.
4. Add generic event emit helpers in JS and Python.
5. Expand fallback ArchiveResult synthesis so on_AfterSnapshot__... hooks behave like on_Snapshot__... hooks.
6. Delete chrome_wait barrier hooks.
7. Move Chrome launch/tab onto real event emission.
8. Define the bg pre-navigation settle semantics needed before moving chrome_navigate to AfterSnapshotChromeTabReady.
9. Move post-navigation extractors to SnapshotChromeTabNavigated.
10. Move clearly artifact-consuming late hooks to AfterSnapshot.
11. Rewrite or keep direct-source URL parsers based on their actual intended inputs.
12. Remove file markers and sibling log polling.
13. Rewrite tests and docs around event semantics and derived.env.

The final design is:

1. one hook syntax: on_<Event>__...
2. hooks emit real facts on stdout
3. hooks may include reserved env patches for durable aggregate state
4. abx-dl routes facts generically
5. abx-dl auto-emits After<Event> when an event subtree settles
6. abx-dl reduces prior scoped events into current key/value context
7. abx-dl mirrors crawl context to CRAWL_DIR/derived.env and snapshot overlays to SNAP_DIR/derived.env only
8. hooks receive reduced context as env vars and current event payload as CLI args
9. SnapshotCleanup waits for the full root Snapshot tree

That preserves the self-healing “read fresh current state” behavior from the old file-based system without keeping the fractured marker-file protocol.

• make root hook-driving lifecycle events use stable string names like CrawlSetup and Snapshot
• dispatch crawl-tree and snapshot-tree hooks dynamically by event string
• route hook-emitted facts generically onto abxbus, including synthetic After<Event> barriers
• reduce prior scoped events into current crawl/snapshot context
• mirror crawl context to CRAWL_DIR/derived.env
• mirror per-snapshot overlays to SNAP_DIR/derived.env without writing snapshot state back to crawl scope
• launch hooks with full reduced context as env vars and current event payload as CLI args
• expand fallback ArchiveResult synthesis so on_AfterSnapshot__... hooks work the same as on_Snapshot__... hooks
• keep SnapshotCleanup waiting for the full root Snapshot tree

• add generic emit_event_record(...) helpers in JS and Python with reserved env patch support
• migrate Chrome launch/tab/navigate to emit real events plus durable env state
• move pre-navigation listeners onto SnapshotChromeTabReady
• move post-navigation extractors onto SnapshotChromeTabNavigated
• move clearly artifact-consuming late hooks onto AfterSnapshot
• review direct-source parsers before moving them late
• remove marker-file protocols like prenav.json and .configured
• remove sibling stdout.log scraping and replace it with real facts / reduced env state
• stop using file existence as readiness
• rewrite tests and docs around emitted events and derived.env

• consume UrlDiscovered instead of relying on hooks emitting child Snapshot records directly
• decide crawl policy for materializing new snapshots from UrlDiscovered
• ensure ArchiveBox’s parallel snapshot execution continues to treat crawl context as upstream-only and snapshot context as local overlay
• update any integration tests or docs that assume marker-file-driven Chrome coordination or old Snapshot discovery records