woods 2.0.0.beta4 → 2.0.0

data/plugin/skills/woods-diagnose/SKILL.md

@@ -36,26 +36,57 @@ bundle exec rails runner 'Rails.application.eager_load!; puts "eager load ok"'
 
 Use the application's normal Docker command and environment variables when applicable. Fix boot/eager-load failures before Woods.
 
+### Watcher setup cannot find already installed Rails or dependencies
+
+Compare normal task discovery with installer preflight in the same container and
+application environment. Early Git builds of the watcher installer stripped
+`BUNDLE_PATH` and `BUNDLE_APP_CONFIG`; Woods `2.0.0` includes the #540 fix.
+Record the loaded revision and bundle configuration source before
+reinstalling dependencies or writing a local bundle-path workaround. Follow the
+[watcher installation guide](https://github.com/lost-in-the/woods/blob/main/docs/WATCH_DAEMON.md#managed-development-startup).
+
+### Puma cannot load the Woods plugin after switching branches
+
+Early generated guards checked only whether Woods was activated. An older gem
+can satisfy that check without providing `puma/plugin/woods.rb`. The #542 fix is
+included in Woods `2.0.0`: verify the loaded revision, then use a supporting
+bundle to preview and apply `bin/rails generate woods:watch --operation update
+--mode puma`. The updated guard checks the active gem's require paths, so older
+gems boot without a watcher. Repeating setup does not upgrade the guard. Follow
+the [owned setup runbook](https://github.com/lost-in-the/woods/blob/main/docs/WATCH_DAEMON.md#ownership-updates-and-removal);
+do not change the owned directive or receipt manually.
+
 ### Watch repeatedly exits 75
 
+First identify the lifecycle owner. The raw task deliberately exits 75; a bare
+Foreman entry then stops all services. Managed `woods-watch`/Puma setup (#538)
+is included in Woods `2.0.0`: verify installed executable/generator help and
+loaded revision before proposing it. Supporting launchers retry boot failures,
+reject idle TTL, and park ownership/protocol conflicts until owner restart.
+Inspect separate supervision state rather than treating its parent PID as a
+healthy daemon. Before the first resolved index path, use launcher logs.
+Do not remove claims or kill PIDs from status to force takeover. See
+[startup diagnosis](https://github.com/lost-in-the/woods/blob/main/docs/TROUBLESHOOTING.md#watcher-startup-or-planned-restart-fails).
+
 Check the installed version's watch guide. Older releases, including
 `2.0.0.beta2`, can rediscover the same restart-trigger paths on every boot. Stop
 the supervisor, run one successful full extraction, then restart the standalone
 watch task. Do not assume automatic startup reconciliation exists in that release.
 For versions documenting environment-boot snapshots, confirm that the command is
-`bundle exec rake woods:watch`, with no preceding `environment` task, and check
+the application's actual `rails woods:watch` or `rake woods:watch` entrypoint,
+with no preceding `environment` task, and check
 whether boot inputs keep changing during initialization or catch-up.
 
 ### Watch retains facts from an initializer deleted while stopped
 
-Record the installed revision. Unreleased after `2.0.0.beta3`, startup preserves
+Record the installed revision. In Woods `2.0.0`, startup preserves
 registered deleted boot inputs as full-extraction obligations. On earlier builds,
 stop watch, run a successful full extraction in a fresh process, then restart
 standalone `woods:watch`. See the installed version's watch guide.
 
 ### A cleaned index directory still exists
 
-Unreleased after `2.0.0.beta3`, `woods:clean` retains the output directory and
+In Woods `2.0.0`, `woods:clean` retains the output directory and
 hidden extraction guard for concurrent writer coordination. Verify published
 artifacts are gone; do not remove that guard while writers may be running.
 
@@ -134,7 +165,7 @@ with `server.version`; missing/null is unknown, not a failure. A validator
 major-version warning calls for full extraction and upgrade review, while a match
 does not certify retained units were migrated. See [writer provenance](https://github.com/lost-in-the/woods/blob/main/docs/PUBLISHED_INDEX.md#manifest-writer-provenance).
 
-Unreleased after `2.0.0.beta3`: incremental/refresh handled source errors keep
+Included in Woods `2.0.0`: incremental/refresh handled source errors keep
 the previous generation active and leave watch batches pending. Repair the
 logged source error and retry the complete batch; see
 [handled source errors](https://github.com/lost-in-the/woods/blob/main/docs/INCREMENTAL_EXTRACTION.md#handled-source-errors-and-retry).
@@ -177,6 +208,24 @@ access with its `WOODS_GIT_DIR` setting. A failed history stream is discarded;
 repair git access and run full extraction to refresh retained metadata. See the
 [history contract](https://github.com/lost-in-the/woods/blob/main/docs/CONFIGURATION_REFERENCE.md#git-enrichment-history).
 
+If per-unit Git metadata is absent, check `git --version` inside the extraction
+process/container as well as repository access. Builds with #551 warn when Git
+cannot execute and a repository is expected; older versions may be silent.
+Source archives without a Git directory remain supported. `GIT_SHA` and
+structural `ready` do not certify history availability. After repairing Git,
+run full extraction; see the
+[missing-executable diagnostic](https://github.com/lost-in-the/woods/blob/main/docs/TROUBLESHOOTING.md#git-executable-is-missing-from-the-extraction-environment).
+
+For linked-worktree provenance/history mismatches, check the installed version's
+`WOODS_GIT_DIR` support and compare the selected branch and exact SHA inside the
+extraction environment. Mount the complete shared `.git` at its original path
+with no override, or select `/mounted-common/worktrees/<id>` within a relocated
+complete mount. Derive `<id>` from Git metadata, not the branch name. Selecting
+the shared root uses the primary checkout's HEAD and also changes incremental
+ranges. A commit alone may leave the source-file watcher idle; run full
+extraction after repair or when current Git history is required. See the
+[worktree mount guide](https://github.com/lost-in-the/woods/blob/main/docs/TROUBLESHOOTING.md#git-directory-mounts-for-linked-worktrees).
+
 After a bundle change or removal of a dynamically defined job, incremental
 extraction can retain stale runtime units. Use a fresh process with the updated
 bundle for full extraction, then validate. For missing external gem paths,
@@ -205,13 +254,13 @@ Compare the client config with the exact command, absolute `cwd`, bundle, and in
 bundle exec woods-mcp-start ./tmp/woods
 ```
 
-If startup says `Could not resolve a published Woods index` (unreleased after
-`2.0.0.beta3`) or names a missing `manifest.json` on older versions, first check
+If startup says `Could not resolve a published Woods index` (included in Woods
+`2.0.0`) or names a missing `manifest.json` on older versions, first check
 the selected index path: an atomic index uses `generation.json` to locate its
 payload manifest. The new headline does not change index validation or recovery.
 Point at an existing index before suggesting a new extraction. Prefer the
-explicit path above; `WOODS_DIR` is also supported. An unreleased change after
-`2.0.0.beta3` adds `WOODS_OUTPUT` after those two choices, so verify the installed
+explicit path above; `WOODS_DIR` is also supported. Woods `2.0.0` includes
+`WOODS_OUTPUT` after those two choices, so verify the installed
 version's configuration guide before relying on that fallback.
 
 Then reconnect through the MCP client and call `woods_status`. Use client-native tool inspection after initialization. Expect 14 packaged Index tools, not all conditional schemas.
@@ -253,6 +302,18 @@ paging alone only visits the discovered prefix. See the
 
 ## 4. Check semantic retrieval
 
+Do not treat structural `ready: true` or bootstrap `hydrated` as proof that
+embeddings exist. Check the installed reader's capabilities: builds with #549
+expose `woods_status.retriever.corpus`, including locally known vector and
+metadata record counts by type. Missing fields or `null` counts mean unknown,
+not zero. Counts include chunks and do not certify complete unit coverage.
+When both stores are known empty, supporting readers return `empty_index` with
+embed or explicit lexical-mode guidance. Follow the
+[corpus diagnostic contract](https://github.com/lost-in-the/woods/blob/main/docs/RETRIEVAL_GUIDE.md#semantic-corpus-diagnostics).
+Older readers need direct embedding-artifact checks; installing this plugin
+does not update the serving gem. Keep reader revision and index writer version
+separate when comparing results.
+
 Configured retrieval defaults (#446) are available in Woods `2.0.0.beta3`. For an installed
 version that supports them, an omitted tool budget uses the serving retriever's
 configured default; an explicit budget overrides it. Standalone MCP does not
@@ -296,6 +357,14 @@ endpoint to silence this warning.
 
 Console failures are live Rails/config/security failures, not Index failures. Verify authorized environment, Rails boot, `WOODS_CONSOLE_CONFIG` or direct `cwd`, blocked-table policy, credentials, and stderr.
 
+For stdio parse errors or response mismatches during tool calls, check for Rails
+logs on stdout. Through Woods `2.0.0.beta4`, stdout is restored after boot;
+configure the Console process's logger to use stderr or a file. Runtime stdout
+isolation is included in Woods `2.0.0`: verify a patched installed revision
+before relying on it. Prefer `bundle exec rake woods:console`; direct Rails
+runner invocation cannot capture output already emitted during Rails boot.
+See the [Console logging diagnosis](https://github.com/lost-in-the/woods/blob/main/docs/CONSOLE_MCP_SETUP.md#rails-logs-break-mcp-protocol).
+
 For MySQL SQL refusals, inspect the executing session's `sql_mode` and the installed version's Console guide. Do not change quote modes to bypass a security refusal.
 
 For SQLite SQL refusals on `2.0.0.beta4` or a reviewed revision containing its Console corrections, consult the installed Console guide for supported identifier and table-reference syntax. Simplify the query to supported syntax; never relax the blocked-table or function policy. These builds also check resolved default scopes and scan normalized response values. Confirm a patched gem is published before recommending it, and check the installed version’s canonical Console guide; do not infer release availability from this plugin.

data/plugin/skills/woods-investigate/SKILL.md

@@ -16,7 +16,7 @@ Follow the [agent guide](https://github.com/lost-in-the/woods/blob/main/docs/AGE
 when instructions are absent. A registered tool does not establish retrieval
 readiness or authorize maintenance or live Console access.
 
-Call `woods_status` before relying on the index. Require a ready index with a current generation and non-zero counts for the types you need; use `codebase_retrieve` only when status reports retrieval enabled. If status is unhealthy or the generation predates the code under review, report that and ask the owner to run `woods:incremental` or `woods:extract` — do not present "not found" as proof the code does not exist.
+Call `woods_status` before relying on the index. Require a ready index with a current generation and non-zero counts for the types you need; verify the retrieval mode and its data before using `codebase_retrieve` (see Conceptual questions below). If status is unhealthy or the generation predates the code under review, report that and ask the owner to run `woods:incremental` or `woods:extract` — do not present "not found" as proof the code does not exist.
 
 ## The default loop
 
@@ -31,9 +31,9 @@ Identifiers are namespaced and typed; never invent one from a filename when `sea
 
 - **Code review / change impact**: `lookup` the changed unit, then `dependents` at depth 1 before going deeper. Group results by relationship type and layer; report direct dependents separately from inferred downstream impact. A graph edge is not test coverage — select tests from mappings and repository search.
 - **Audit / architecture assessment**: `graph_analysis` for orphans, dead ends, hubs, cycles, bridges, cross-database edges, volatile dependencies, and undeclared package edges; `domain_clusters` for architectural domains; `pagerank` for high-impact units worth reading first.
-- **Investigating behavior / debugging**: `trace_flow` from the user-visible entry point (route, controller action, job, mailer, service), `lookup` at ambiguous steps, and verify anything conditional or dynamically dispatched in source and tests — do not infer call order from a dependency edge.
+- **Investigating behavior / debugging**: find the exact indexed unit with `search` and `lookup`, then use `trace_flow` with `UnitIdentifier` or `UnitIdentifier#method` (for example, `CheckoutService#order`). Bare `order` names a unit, potentially a factory, rather than locating an application method. Receiverless local calls may remain unexpanded; inspect their source or trace the owning unit's method explicitly. Flow output is not proof of runtime execution or exhaustive call coverage. See the [flow workflow](https://github.com/lost-in-the/woods/blob/main/docs/AGENT_GUIDE.md#trace-a-feature-flow).
 - **Onboarding**: `structure` for the codebase overview, `lookup` and `dependencies`/`dependents` for a unit's neighborhood, and `domain_clusters` for the domain map, then the default loop on the units that matter.
-- **Conceptual questions**: `codebase_retrieve` when status says ready; govern with `budget` (never `limit`), then verify key units with `lookup`.
+- **Conceptual questions**: check retrieval mode and data before `codebase_retrieve`; structural `ready` alone does not establish semantic availability. On readers supporting #549, inspect `retriever.corpus`; missing or unknown counts require checking embedding artifacts. Govern with `budget` (never `limit`), then verify key units with `lookup`.
 
 ## Boundaries
 
@@ -60,12 +60,12 @@ PORO and library targets. No dependents or test-only dependents do not establish
 absence of production callers; check source before making that claim.
 
 The response `graph_coverage` notice, `total_is_exact` field, and human label
-`witness types unambiguous` (#470/#471) are unreleased after Woods `2.0.0.beta3`.
+`witness types unambiguous` (#470/#471) are included in Woods `2.0.0`.
 Verify the installed server version and actual response fields; this plugin does
 not add them. Apply these limits to older servers even without the notice.
 Supporting stdio and HTTP servers expose the paginated traversal payload in
 `structuredContent.data` independently of the text renderer (#481, also
-unreleased after `2.0.0.beta3`). Check the installed response; older default
+included in Woods `2.0.0`). Check the installed response; older default
 responses may carry only text. Do not pass an unsupported `format` argument.
 
 `total_is_exact: false` means a budget-limited prefix; a true value describes only
@@ -106,7 +106,7 @@ recorded reachability does not establish observed execution. See the
 
 Pass explicit `limit` and `offset` when paging `graph_analysis`. Enforcing the
 advertised default of 20 rows per section and preserving total/offset on last
-and empty pages (#519) are unreleased after `2.0.0.beta3`; check the installed
+and empty pages (#519) are included in Woods `2.0.0`; check the installed
 response rather than inferring support from the plugin version. On supporting
 servers, read `<section>_total` and `<section>_offset` in JSON, or the human
 pagination notice. An empty later page does not mean no findings. Totals count

data/plugin/skills/woods-mcp-config/SKILL.md

@@ -55,11 +55,26 @@ when unavailable. See the
 
 Use this shape for any stdio-capable MCP client, adapted to the client's configuration location. `woods-mcp-start` validates and launches; it does not install or auto-restart.
 
+MCP registration does not start automatic indexing. Existing readers observe
+published generations without reconnecting; separately verify the watcher owner,
+startup catch-up, and a real edit. Native launcher/Puma installation (#538) is
+included in Woods `2.0.0`: check installed `woods-watch` and generator help
+before offering it. Preserve the existing external service in Docker/Grove and
+keep its source/index aligned across switches. See
+[automatic maintenance](https://github.com/lost-in-the/woods/blob/main/docs/AUTOMATIC_MAINTENANCE.md).
+
 Writer-version provenance (#323) is available in Woods `2.0.0.beta3`; check the installed
 gem version's release notes before expecting `index.woods_version` in `woods_status`. It reports
 the last manifest publisher, independently of `server.version`. Treat missing/null
 as unknown and see [writer provenance](https://github.com/lost-in-the/woods/blob/main/docs/PUBLISHED_INDEX.md#manifest-writer-provenance).
 
+Verify semantic retrieval separately from structural `ready`. A reachable
+provider and bootstrap `hydrated` can coexist with empty stores. If the recorded
+reader supports #549, inspect `retriever.corpus` for local record counts and
+known-empty diagnostics; absent or unknown counts require checking embedding
+artifacts. These fields do not certify embedding coverage. See the
+[readiness distinction](https://github.com/lost-in-the/woods/blob/main/docs/RETRIEVAL_GUIDE.md#semantic-corpus-diagnostics).
+
 When Woods is installed only in Docker, prefer running the server through the application container:
 
 ```json
@@ -74,7 +89,13 @@ When Woods is installed only in Docker, prefer running the server through the ap
 }
 ```
 
-Use a host-side bundle only after verifying Ruby, the application bundle, and the index are available on the host. Always pass the path visible to the process that runs `woods-mcp`. Prefer an explicit index path on all versions. After `2.0.0.beta3`, an unreleased change adds `WOODS_OUTPUT` as a fallback after the positional path and `WOODS_DIR`; check the installed version's configuration guide before relying on it. `woods-mcp-start` still refuses a missing path rather than selecting its working directory.
+Use a host-side bundle only after verifying Ruby, the application bundle, and the index are available on the host. Always pass the path visible to the process that runs `woods-mcp`. Prefer an explicit index path on all versions. Woods `2.0.0` includes `WOODS_OUTPUT` as a fallback after the positional path and `WOODS_DIR`; check the installed version's configuration guide before relying on it. `woods-mcp-start` still refuses a missing path rather than selecting its working directory.
+
+For linked worktrees, verify source/index alignment and the extraction's Git
+branch and exact SHA. Preserve the complete shared Git layout and select the
+worktree-specific directory when using the installed version's `WOODS_GIT_DIR`
+override; the shared root selects the primary checkout's HEAD. Follow the
+[worktree mount guide](https://github.com/lost-in-the/woods/blob/main/docs/TROUBLESHOOTING.md#git-directory-mounts-for-linked-worktrees).
 
 A read-only index mount is sufficient for structural tools. The `reload` tool for in-memory semantic retrieval also takes Woods' shared on-disk writer lock, so the MCP process needs write access to the index directory. Without it, reload returns a typed degraded error and keeps serving the previous aligned generation. Either grant that access or restart the MCP process after publishing a new embedded index.
 
@@ -116,6 +137,12 @@ Then add a direct Console process:
 
 For Docker/SSH, configure `~/.woods/console.yml` or `WOODS_CONSOLE_CONFIG`; the launcher owns process replacement. Direct Docker stdio uses `docker exec -i`, or `docker compose exec -T` to disable Compose's pseudo-TTY while retaining stdin.
 
+Reserve stdout for MCP. Through Woods `2.0.0.beta4`, configure the Console
+process's Rails logger to use stderr or a file, including logs during queries.
+Runtime stdout isolation is included in Woods `2.0.0`; verify the installed
+revision before relying on it. Use the rake entry point to capture Rails boot
+output as well.
+
 Console registers nine default tools. `config.console_embedded_read_tools = true` explicitly adds `console_sql` and `console_query` for eleven total. Tier 2, Tier 3, and `console_eval` are inventory-only in supported packaged modes.
 
 ## Shape 3: Authenticated Console HTTP

data/plugin/skills/woods-setup/SKILL.md

@@ -38,7 +38,7 @@ This skill describes the Woods 2.x line; the authoritative minimum version lives
 
 Choose the published version using the canonical
 [installation guide](https://github.com/lost-in-the/woods/blob/main/docs/GETTING_STARTED.md#1-install-the-gem)
-and its linked README release table. When only prereleases are published for
+and confirm the selected version on RubyGems. When only prereleases are published for
 2.x, add the exact published beta/RC constraint shown there to the development
 group; `~> 2.0` does not select prereleases. Use `gem "woods", "~> 2.0"` only after
 a stable 2.x release is published. Follow the selected version's tag docs and
@@ -108,9 +108,58 @@ When Woods is installed only in Docker, launch it through the application servic
 }
 ```
 
+For linked worktrees, verify source/index alignment and the extraction's Git
+branch and exact SHA. Preserve the complete shared Git layout and select the
+worktree-specific directory when using the installed version's `WOODS_GIT_DIR`
+override; the shared root selects the primary checkout's HEAD. Follow the
+[worktree mount guide](https://github.com/lost-in-the/woods/blob/main/docs/TROUBLESHOOTING.md#git-directory-mounts-for-linked-worktrees).
+
 Reconnect and call `woods_status`, then `search`, `lookup`, and `dependents` for a known class. The normal Index Server has 14 tools. `codebase_retrieve` requires configured embeddings in semantic mode; see the lexical capability check below for the opt-in provider-free mode.
 
-Offer to add `bundle exec rake woods:watch` to the existing development process manager. When authorized, it catches up missed changes and automatically maintains the structural index; the Index Server refreshes on its next call, so ordinary edits need no manual extraction or MCP restart. Use the standalone watch command; do not prepend the `environment` task. Check the installed version's watch guide before relying on automatic startup reconciliation. State that live boot-captured changes require supervisor restart, Docker may need `WOODS_WATCH_POLL=1`, and semantic vectors still need `woods:embed_incremental`.
+Structural readiness alone does not verify semantic retrieval. If that mode is
+requested, confirm an embedding run and a useful retrieval result. Readers with
+#549 expose `retriever.corpus` counts; check the installed capability before
+expecting them. Unknown counts are not zero, and positive counts do not certify
+complete embedding coverage. See
+[corpus diagnostics](https://github.com/lost-in-the/woods/blob/main/docs/RETRIEVAL_GUIDE.md#semantic-corpus-diagnostics).
+
+Offer one automatic-maintenance owner within the setup scope. The managed launcher,
+watcher generator, and Puma adapter (#538) are **included in Woods `2.0.0`**.
+Record the loaded gem path and revision, then verify `bundle exec woods-watch
+--help` and `bin/rails generate woods:watch --help` before using them. Follow the
+[managed startup runbook](https://github.com/lost-in-the/woods/blob/main/docs/WATCH_DAEMON.md#managed-development-startup):
+Puma for simple Rails startup, an explicit verified Foreman command/Procfile, or
+the existing external Docker/Grove supervisor. Preview before applying within
+existing authorization. Preserve `bin/dev`; never claim an unused Procfile is
+active. Keep the portable receipt with generated files and respect edit conflicts.
+Puma installation supports the normal default `config/puma.rb` route; an existing
+environment-specific file or custom `-C` route needs a different explicit owner.
+The #542 guard included in Woods `2.0.0` checks plugin files in the active gem, including Git/path
+bundles. Older gems without the plugin skip watcher startup. Existing generated
+Puma setups need an explicit `--operation update --mode puma` (preview first) to
+upgrade the owned guard in place; repeating setup preserves it. The wrapper and
+Foreman entry still require a supporting gem. Remove owned setup before a permanent
+downgrade; do not hand-edit the receipt or its managed block.
+Run setup in the normal application bundle environment. The #540 fix in Woods `2.0.0`
+preserves `BUNDLE_PATH`, `BUNDLE_APP_CONFIG`, and group selection during preflight;
+older Git builds may falsely report missing gems. Check the loaded revision
+before changing persistent Bundler settings to work around that error.
+The #544 fix in Woods `2.0.0` makes generator refusals exit nonzero. Earlier Git builds
+can exit zero after printing a refusal; verify the diagnostic and applied setup
+before treating their exit status as installation success.
+For an interrupted install, use the generator's `--operation recover --pretend`
+before applying recovery; preserve journals when concurrent edits block it. The
+Rails generator command boots the app first: use the runbook's direct bundled
+Ruby recovery helper when an initializer prevents boot.
+
+On older gems, use the raw task only with a restart-capable external supervisor;
+do not place it bare in Foreman, where exit 75 stops the whole stack. Use the
+application's real Rails task entrypoint without a preceding `environment` task.
+Managed mode rejects idle TTL and does not take over conflicting owners. Verify
+startup catch-up, an edit, and a planned restart through the existing MCP reader;
+for Grove, also verify worktree/source/index alignment. Docker may need polling.
+Semantic vectors still need `woods:embed_incremental`; hooks and MCP registration
+do not install watcher startup.
 
 Foreign-host heartbeat trust (`WOODS_WATCH_TRUST_FOREIGN_HOST=1`, #321) is available in
 Woods `2.0.0.beta3`.
@@ -155,13 +204,18 @@ for transport, retry and custom-root limits.
 
 ## Ask before expanding scope
 
-For pgvector, match the provider output and migration dimensions within 1–2,000. Default `text-embedding-3-large` output (3,072) needs an explicit smaller provider width or another backend; never silently truncate vectors. Early adapter/generator refusal is unreleased after `2.0.0.beta3`, so check the installed revision. See the [dimension contract](https://github.com/lost-in-the/woods/blob/main/docs/CONFIGURATION_REFERENCE.md#pgvector-postgresql).
+For pgvector, match the provider output and migration dimensions within 1–2,000. Default `text-embedding-3-large` output (3,072) needs an explicit smaller provider width or another backend; never silently truncate vectors. Early adapter/generator refusal is included in Woods `2.0.0`, so check the installed revision. See the [dimension contract](https://github.com/lost-in-the/woods/blob/main/docs/CONFIGURATION_REFERENCE.md#pgvector-postgresql).
 
 Require explicit approval before adding Ollama/OpenAI, pgvector/Qdrant, secrets, Console MCP/live-data access, HTTP transport, or purge overrides. The `:local` preset avoids cloud keys but requires the `sqlite3` gem, an installed/running Ollama service, and a pulled model (`ollama pull nomic-embed-text` by default); `:shared_filesystem` avoids sqlite3 but still uses Ollama. Recommend `gem "tokenizers", "~> 0.5"` for exact counting on dense Ruby source, while stating that it is optional.
 
 ## Handoff
 
-Report the Woods version, branch, files changed, commands/results, index path, MCP calls verified, semantic retrieval status, Console status, and unresolved risks. Never infer availability from source schemas alone.
+Report the Woods version/revision, branch, files changed, commands/results, index
+path, MCP calls verified, semantic retrieval and Console status, and unresolved
+risks. For automatic maintenance record the owner, actual startup command,
+completed catch-up, observed edit/restart, and worktree verification. Report
+refresh hooks, session checks, and context hints separately. Never infer
+availability from source schemas or a live process alone.
 
 Canonical runbook: [AGENT_SETUP.md](https://github.com/lost-in-the/woods/blob/main/docs/AGENT_SETUP.md).
 

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: woods
 version: !ruby/object:Gem::Version
-  version: 2.0.0.beta4
+  version: 2.0.0
 platform: ruby
 authors:
 - Leah Armstrong
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-09-22 00:00:00.000000000 Z
+date: 2026-09-23 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: mcp
@@ -113,6 +113,7 @@ description: |
 email:
 - info@leah.wtf
 executables:
+- woods-watch
 - woods-mcp
 - woods-mcp-start
 - woods-console-mcp
@@ -133,6 +134,7 @@ files:
 - assets/woods-wordmark-white-with-bg.png
 - docs/AGENT_GUIDE.md
 - docs/AGENT_SETUP.md
+- docs/AUTOMATIC_MAINTENANCE.md
 - docs/BACKEND_MATRIX.md
 - docs/CLIENT_HOOKS.md
 - docs/CONFIGURATION_REFERENCE.md
@@ -171,11 +173,14 @@ files:
 - exe/woods-mcp
 - exe/woods-mcp-http
 - exe/woods-mcp-start
+- exe/woods-watch
 - lib/generators/woods/install_generator.rb
 - lib/generators/woods/pgvector_generator.rb
 - lib/generators/woods/templates/add_pgvector_to_woods.rb.erb
 - lib/generators/woods/templates/create_woods_tables.rb.erb
 - lib/generators/woods/templates/woods.rb.tt
+- lib/generators/woods/watch_generator.rb
+- lib/puma/plugin/woods.rb
 - lib/tasks/woods.rake
 - lib/tasks/woods_checks.rake
 - lib/tasks/woods_evaluation.rake
@@ -235,6 +240,7 @@ files:
 - lib/woods/console/sql_table_scanner.rb
 - lib/woods/console/sql_validator.rb
 - lib/woods/console/sqlite_read_guard.rb
+- lib/woods/console/stdio_transport.rb
 - lib/woods/console/table_gate.rb
 - lib/woods/console/tool_specs.rb
 - lib/woods/console/tools/tier1.rb
@@ -430,6 +436,7 @@ files:
 - lib/woods/resilience/retryable_provider.rb
 - lib/woods/resolved_config.rb
 - lib/woods/retrieval/context_assembler.rb
+- lib/woods/retrieval/corpus_status.rb
 - lib/woods/retrieval/lexical_assembler.rb
 - lib/woods/retrieval/lexical_index.rb
 - lib/woods/retrieval/query_classifier.rb
@@ -469,6 +476,7 @@ files:
 - lib/woods/source_inputs/verifier.rb
 - lib/woods/storage/graph_store.rb
 - lib/woods/storage/inapplicable_backend.rb
+- lib/woods/storage/local_corpus_stats.rb
 - lib/woods/storage/metadata_store.rb
 - lib/woods/storage/pgvector.rb
 - lib/woods/storage/qdrant.rb
@@ -491,10 +499,32 @@ files:
 - lib/woods/util/uuid5.rb
 - lib/woods/version.rb
 - lib/woods/watch/boot_snapshot.rb
+- lib/woods/watch/child_environment.rb
+- lib/woods/watch/cli.rb
 - lib/woods/watch/daemon.rb
+- lib/woods/watch/event_stream.rb
+- lib/woods/watch/guardian.rb
+- lib/woods/watch/installation.rb
+- lib/woods/watch/installation/layout.rb
+- lib/woods/watch/installation/options.rb
+- lib/woods/watch/installation/planner.rb
+- lib/woods/watch/installation/probe.rb
+- lib/woods/watch/installation/receipt.rb
+- lib/woods/watch/installation/recovery.rb
+- lib/woods/watch/installation/templates.rb
+- lib/woods/watch/lifecycle.rb
 - lib/woods/watch/listen_watcher.rb
+- lib/woods/watch/managed_child.rb
+- lib/woods/watch/managed_cleanup.rb
+- lib/woods/watch/managed_process.rb
 - lib/woods/watch/polling_watcher.rb
+- lib/woods/watch/puma_adapter.rb
+- lib/woods/watch/puma_child.rb
 - lib/woods/watch/status.rb
+- lib/woods/watch/supervision_records.rb
+- lib/woods/watch/supervision_status.rb
+- lib/woods/watch/supervisor.rb
+- lib/woods/watch/supervisor_reporting.rb
 - lib/woods/watch/tree_scan.rb
 - lib/woods/watch/watcher.rb
 - plugin/.claude-plugin/plugin.json
@@ -517,10 +547,10 @@ licenses:
 - MIT
 metadata:
   homepage_uri: https://github.com/lost-in-the/woods
-  source_code_uri: https://github.com/lost-in-the/woods/tree/v2.0.0.beta4
-  changelog_uri: https://github.com/lost-in-the/woods/blob/v2.0.0.beta4/CHANGELOG.md
+  source_code_uri: https://github.com/lost-in-the/woods/tree/v2.0.0
+  changelog_uri: https://github.com/lost-in-the/woods/blob/v2.0.0/CHANGELOG.md
   bug_tracker_uri: https://github.com/lost-in-the/woods/issues
-  documentation_uri: https://github.com/lost-in-the/woods/tree/v2.0.0.beta4/docs
+  documentation_uri: https://github.com/lost-in-the/woods/tree/v2.0.0/docs
   rubygems_mfa_required: 'true'
 post_install_message:
 rdoc_options: []