thumbgate 1.35.0 → 1.37.2

package/.agents/skills/cobble-hot-store-compare-not-clone/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: cobble-hot-store-compare-not-clone
+description: >
+  CobbleDB (Perplexity 2026-09) is a batch-read hot store, not a ThumbGate
+  clone. Steal the three-plane FORMAT (durable state / batched delivery /
+  query-time MultiGet + hedge + hot subset) onto existing lesson rails; never
+  vendor RocksDB, YTsaurus, Pillar, or Lorry. Slash: /cobble-hot-store-compare-not-clone.
+---
+
+# CobbleDB — compare, do not clone
+
+## Goal
+
+Produce fail-closed honesty for whom: ThumbGate agents writing feedback or
+serving lessons — so processing never contends with PreToolUse reads, batch
+lookups hedge slow sources, and only promoted/matchable records occupy the
+hot path.
+
+## Constraints
+
+| NEVER | ALWAYS |
+| --- | --- |
+| Clone CobbleDB / Pillar / Lorry / RocksDB / YTsaurus | Map planes onto existing lesson rails |
+| Write processing output onto `hybrid-feedback-context` | Durable log → `feedback-to-memory` → retrieval |
+| Sequential per-key hot reads of a batch | `batchGetPrepared` (partition + hedge) |
+| Dump raw `feedback-log.jsonl` into retrieval | `promoted_matchable` (or `fresh_high_value`) subset |
+| Overwrite embedding/chunk versions | Coexist representations |
+| Dual-edit DIRTY lesson-graph PR #3650 | Leave graph layer alone |
+| Quote Perplexity 5× / 20% as ours | Measure ThumbGate rails with `perf-budget-check` |
+| Hero Continuity / net-new storage SKU | ECI: existing retrieval rails only |
+
+HARD fail closed. REFUSE SKU clones.
+
+## Reference
+
+- https://www.perplexity.ai/hub/blog/cobbledb
+- `scripts/cobble-hot-store-split.js`
+- `scripts/lesson-retrieval.js` · `scripts/feedback-to-memory.js` · `scripts/memory-vs-rag-route.js`
+- `docs/agents/cobble-hot-store-split.md`
+- `/high-roi-steal-and-finish` · `/eci-thumbgate-ip-wall`
+
+## Examples (show, don't tell)
+
+Weak: Summarize CobbleDB and add a RocksDB cluster class.
+
+Gold:
+
+```bash
+$ npx thumbgate cobble-hot-store-split --json
+ok: true
+status: ready
+$ node --test tests/cobble-hot-store-split.test.js
+# processing → hot write → coupled_processing_hot_write deny
+```
+
+## Procedures
+
+```bash
+npx thumbgate cobble-hot-store-split --json
+npx thumbgate cobble-hot-store-split --map-only
+npx thumbgate cobble-hot-store-split --trace=tests/fixtures/cobble-hot-store-split-coupled.json --json
+npx thumbgate cobble-hot-store-split --keys=lesson-1,lesson-2 --json
+npm run test:cobble-hot-store-split
+```
+
+1. Classify every write as `durable`, `delivery`, or `hot`.
+2. Deny processing → hot coupling.
+3. Require batched MultiGet (+ hedge) for ≥3 serving keys.
+4. Keep the hot subset to promoted/matchable records.
+5. Refuse `--clone-cobbledb`.
+
+## Rubric
+
+- gold default (this repo) → `ok=true`, `status=ready`
+- `--trace` coupled fixture → `ok=false`, `coupled_processing_hot_write`
+- `--clone-cobbledb` → `ok=false`, `cobble_clone_refused`
+- sequential ≥3-key hot read → `sequential_hot_batch`
+- unbounded hot subset → `unbounded_hot_subset`
+- doctor: `npm run test:cobble-hot-store-split` PASS
+- evidence: command output in the same turn

package/.agents/skills/colab-compute-honesty-not-clone/SKILL.md

@@ -0,0 +1,70 @@
+---
+name: colab-compute-honesty-not-clone
+description: >
+  Google Colab /signup is a compute-unit storefront, not a ThumbGate GPU SKU.
+  Steal CU-pack honesty (subscribe≠receipt, CU≠dedicated GPU, 24h background
+  is Pro+). Never buy Pro/Pro+/PAYG from an agent session. Never clone
+  colab-cli. Slash: /colab-compute-honesty-not-clone.
+---
+
+# Colab compute honesty — compare, do not clone
+
+## Goal
+
+Fail-closed compute claims for whom: ThumbGate evals and fleet offload talk —
+so "Colab Pro+ A100" cannot pass without a live Current-plan receipt.
+Colab stays a neighbor notebook host, not a product.
+
+## Constraints
+
+| NEVER | ALWAYS |
+| --- | --- |
+| Buy Colab Pro / Pro+ / Compute Units | `npx thumbgate colab-compute-honesty --json` |
+| Claim Pro+ because Subscribe is on /signup | `--plan-proof` from a Current-plan receipt |
+| colab-cli / ngrok SSH / zero-cost A100 | GitHub Actions remains the eval runner |
+| Dual-edit unverified `google-colab-pro-runner` theater | Treat that skill as a claim to audit |
+
+HARD fail closed. REFUSE SKU clones. Do NOT spend.
+
+## Reference
+
+- https://colab.research.google.com/signup (live 2026-09-17, `iganapolsky@gmail.com`)
+- PAYG $9.99/100 CU and $49.99/500 CU (buttons disabled on that account)
+- Pro $9.99/mo and Pro+ $49.99/mo still showed Subscribe
+- `scripts/colab-compute-honesty.js`
+
+## Examples (show, don't tell)
+
+Weak: Subscribe to Pro+ and offload gate-eval to A100 via colab-cli.
+
+Gold:
+
+```bash
+$ npx thumbgate colab-compute-honesty --json --map-only
+ok: true
+$ npx thumbgate colab-compute-honesty --json --claim='Colab Pro+ A100 eval sweep'
+ok: false  # paid_feature_without_plan_proof
+```
+
+## Procedures
+
+```bash
+npx thumbgate colab-compute-honesty --json --map-only
+npx thumbgate colab-compute-honesty --json --claim='offload to Colab A100'
+npx thumbgate colab-compute-honesty --json --clone-colab
+npm run test:colab-compute-honesty
+```
+
+1. Open /signup signed in. Subscribe visible ⇒ not a receipt.
+2. Do not click Buy / Subscribe.
+3. Run the doctor on any Colab/GPU offload claim.
+4. Keep ThumbGate tests on GitHub Actions.
+
+## Rubric
+
+- `--map-only` → `ok=true`
+- Pro+/A100/24h without `--plan-proof` → fail
+- live snapshot with Subscribe + `--plan-proof=proplus` → `subscribe_button_is_not_receipt`
+- `--buy-pro` / `colab-cli` → fail
+- doctor tests PASS
+- evidence: `--json` in the same turn

package/.agents/skills/cyberstrike-compare-not-clone/SKILL.md

@@ -0,0 +1,36 @@
+---
+name: cyberstrike-compare-not-clone
+description: >
+  CyberStrikeAI (Ed1s0nZ) turns NL intent into governed cybersecurity actions
+  with HITL + RAG. Steal the FORMAT onto ThumbGate rails; never clone
+  CyberStrike, Eino, WebShell, or C2. Slash: /cyberstrike-compare-not-clone.
+---
+
+# CyberStrikeAI — compare, do not clone
+
+## When
+CyberStrikeAI, Ed1s0nZ, tom_doerr Eino agents, intent becomes governed execution,
+evidence becomes operational memory, human oversight pentest agent.
+
+## Do
+```bash
+npx thumbgate intent-governed-execution --intent="<what to do>" --json
+npx thumbgate intent-governed-execution --map-only
+```
+
+Map FORMAT onto existing rails:
+- **classify** → intent-router + harness-selector
+- **authorize** → session-lease / task-scope; offensive needs `THUMBGATE_ALLOW_OFFENSIVE=1`
+- **gate** → PreToolUse / gates-engine / harness JSON
+- **hitl** → admin-override / protectedApprovals (`--approved` only after a real human grant)
+- **execute** → subagent-profiles + maxChars result governance
+- **evidence** → capture-feedback → lessons → prevention rules
+
+## Never
+- Clone/vendor CyberStrikeAI or CloudWeGo Eino into ThumbGate
+- Ship WebShell / C2 / metasploit recipe packs
+- Invent automation ROI percentages
+- Run offensive tools without authorization grant + HITL
+
+## Source
+https://github.com/Ed1s0nZ/CyberStrikeAI · https://x.com/tom_doerr/status/2094509419929174170 — FORMAT only; not affiliated.

package/.agents/skills/gitlab-sandbox-allowlist-not-trust/SKILL.md

@@ -0,0 +1,77 @@
+---
+name: gitlab-sandbox-allowlist-not-trust
+description: >
+  GitLab 2026-09 FORMAT steal: sandbox allowlists are hops, not trust
+  boundaries. Audit package-registry/proxy/HF destinations and trust-handoff
+  writes on existing ThumbGate egress rails. Never clone GitLab Duo.
+  Slash: /gitlab-sandbox-allowlist-not-trust.
+---
+
+# GitLab sandbox allowlist ≠ trust
+
+## Goal
+
+Produce fail-closed honesty for whom: ThumbGate agents treating
+`deny-network-egress` / `agent-egress-policy` allowlists as sandbox
+containment — so a permitted package proxy cannot silently become the
+escape hop GitLab documented.
+
+## Constraints
+
+| NEVER | ALWAYS |
+| --- | --- |
+| Clone GitLab Duo Agent Platform / execution sandbox SKU | Classify registry/proxy/HF hosts as `bridge` |
+| Treat allowlisted `registry.npmjs.org` as trusted | Independent auth before credentials ride a hop |
+| Auto-promote observed proxy traffic onto `allowHosts` | Put bridge hosts on `bridgeHosts` only |
+| Claim hook/CI/MCP writes are sandbox-contained | Call them trust-handoff |
+| Dual-edit untracked `scripts/gitlab-ai-sandbox-integration.js` | Use this doctor + existing egress policy |
+| Hero Continuity / net-new governance SKU | Map onto PreToolUse + `agent-egress-policy` |
+
+HARD fail closed. REFUSE SKU clones. ECI: maintenance of existing egress rails only.
+
+## Reference
+
+- https://www.infoq.com/news/2026/09/gitlab-ai-sandbox-access/
+- https://about.gitlab.com/blog/ai-agent-sandbox/
+- `scripts/allowlist-bridge-honesty.js`
+- `scripts/agent-egress-policy.js` (`STATIC_ALLOW_BRIDGE`, `ALLOWLIST_BRIDGE_CREDENTIAL`)
+- `THREAT_MODEL.md` policy-vs-containment
+- `/high-roi-steal-and-finish` · `/eci-thumbgate-ip-wall`
+
+## Examples (show, don't tell)
+
+Weak: Summarize GitLab Duo and add a container sandbox class.
+
+Gold:
+
+```bash
+$ npx thumbgate@1.36.1 allowlist-bridge-honesty --json
+allowlistIsTrustBoundary: false
+$ node --test tests/allowlist-bridge-honesty.test.js
+# Authorization to registry.npmjs.org → ALLOWLIST_BRIDGE_CREDENTIAL deny
+```
+
+## Procedures
+
+```bash
+npx thumbgate@1.36.1 allowlist-bridge-honesty --json
+npx thumbgate@1.36.1 allowlist-bridge-honesty --allow-hosts=registry.npmjs.org --treat-allowlist-as-trust --json
+npx thumbgate@1.36.1 allowlist-bridge-honesty --write=.github/workflows/ci.yml --claimed-contained --json
+npm run test:allowlist-bridge-honesty
+```
+
+1. Classify every allowlisted host (`bridge` vs `public` vs `ssrf_private`).
+2. Deny credentialed requests to bridge hops unless `independentAuth` is already granted.
+3. Keep observe→draft from promoting package proxies onto `allowHosts`.
+4. Label privileged-consumer writes as trust-handoff.
+5. Refuse `--clone-gitlab-duo`.
+
+## Rubric
+
+- gold fixture (this repo default gate) → `ok=true`, `allowlistIsTrustBoundary=false`
+- `--treat-allowlist-as-trust` with npm registry → `ok=false`
+- `--clone-gitlab-duo` → `ok=false`
+- observe draft of `registry.npmjs.org` → not in `allowHosts`
+- Authorization to allowlisted npm → `ALLOWLIST_BRIDGE_CREDENTIAL`
+- doctor: `npm run test:allowlist-bridge-honesty` PASS
+- evidence: command output in the same turn

package/.agents/skills/jit-harness-compare-not-clone/SKILL.md

@@ -0,0 +1,34 @@
+---
+name: jit-harness-compare-not-clone
+description: >
+  JIT-Agent (arXiv:2608.25593) is a harness-intelligence model, not a ThumbGate
+  clone. Steal the four-module FORMAT (memory/planning/action/capability) onto
+  existing rails; never train or download JIT-Agent. Slash: /jit-harness-compare-not-clone.
+---
+
+# JIT-Agent — compare, do not clone
+
+## When
+JIT-Agent, arxiv 2608.25593, Rohan Paul harness intelligence, Model-as-a-Harness,
+just-in-time harness evolution, HarnessFactory, four-module agent harness.
+
+## Do
+```bash
+npx thumbgate jit-harness-compose --task="<what the agent should do>" --json
+npx thumbgate jit-harness-compose --map-only
+```
+
+Map modules to existing rails:
+- **memory** → lesson-retrieval / feedback / contextfs
+- **planning** → `bin/agent-loop` Plan + GSD
+- **action** → PreToolUse gates + `harness-selector` + `switchyard-router`
+- **capability** → `subagent-profiles` + MCP allowlists + skills + model pool
+
+## Never
+- Train, serve, or download JIT-Agent-27B / `huggingface.co/JIT-Agent`
+- Clone `bingreeky/JIT` HarnessFactory into ThumbGate
+- Emit free-form harness programs as a new SKU
+- Claim ThumbGate is JIT-Agent
+
+## Source
+https://arxiv.org/abs/2608.25593 · https://github.com/bingreeky/JIT — FORMAT only; not affiliated.

package/.agents/skills/openui-catalog-compose-honesty/SKILL.md

@@ -0,0 +1,64 @@
+---
+name: openui-catalog-compose-honesty
+description: >
+  OpenUI FORMAT steal: catalog-compose-only, root-first streaming, repair-before-claim.
+  Audit agent-composed UI against an allowlisted component catalog; never clone OpenUI,
+  Thesys Gateway, or a generative-UI SKU. Slash: /openui-catalog-compose-honesty.
+---
+
+# OpenUI Catalog-Compose Honesty
+
+## Goal
+Keep agent-composed UI honest: only allowlisted components, root-first streams, repair before any ready/done claim. Steal OpenUI FORMAT — do not clone the product.
+
+## Constraints
+- Never install `@openuidev/cli`, OpenUI Gateway, or Thesys Observability as ThumbGate
+- Never invent components outside the catalog or `eval` / `new Function` generated UI
+- Never claim UI/compose ready without `--repair --claim-ready` evidence
+- Never dual-edit a sibling OpenUI/generative-UI WIP PR
+- ECI: no net-new generative-UI SKU
+
+## Reference
+- https://www.openui.com/ (source FORMAT)
+- Gates: `require-catalog-compose-only`, `require-repair-before-compose-claim`
+- Script: `scripts/openui-catalog-compose-honesty.js`
+- Docs: `docs/agents/openui-catalog-compose-honesty.md`
+
+## Examples
+```bash
+# Clean catalog stream → ready
+npx thumbgate openui-catalog-compose-honesty \
+  --catalog=catalog.json \
+  --stream=compose.txt \
+  --repair --claim-ready --json
+
+# Unknown component / missing root → fail
+npx thumbgate openui-catalog-compose-honesty \
+  --catalog='{"components":["Stack"]}' \
+  --stream=$'header = Ghost()\n' \
+  --json
+```
+
+Catalog:
+```json
+{ "components": ["Stack", "Card", "Table"] }
+```
+
+Stream (root first):
+```
+root = Stack([header, body])
+header = Card()
+body = Table([row])
+```
+
+## Procedures
+1. Check sibling WIP for OpenUI/generative-UI PRs — do not duplicate.
+2. Write or load a component catalog JSON.
+3. Run doctor with `--catalog` + `--stream` (+ `--repair`).
+4. Before any UI/compose done claim, re-run with `--claim-ready --strict`.
+5. Wire gates from `config/gate-templates.json` Agent Honesty pair.
+
+## Rubric
+- ok=true when doctor status is `ready` with `repairClean=true` on the claimed stream
+- ok=false on unknown components, missing root, arbitrary-code markers, or OpenUI SKU clone signals
+- evidence: `--json` report in the same turn

package/.agents/skills/token-shunt-honesty-not-clone/SKILL.md

@@ -0,0 +1,67 @@
+---
+name: token-shunt-honesty-not-clone
+description: >
+  Steal Spotify Portal/shunt FORMAT: PreToolUse blocks untargeted Read above
+  350 lines, bare cat of large files, and full-file dumps. Do NOT install
+  shunt@portal, buy Portal, or claim 90% savings from GitHub App 162279530.
+  Slash: /token-shunt-honesty-not-clone.
+---
+
+# Token shunt FORMAT — intercept, don't clone Portal
+
+## Goal
+
+Produce fail-closed bulk-read intercepts for whom: ThumbGate agents — so
+untargeted Read/cat of large files never dump into the frontier context, and
+Spotify Portal stays a compare-not-clone neighbor, not a SKU or cash rail.
+
+## Constraints
+
+| NEVER | ALWAYS |
+| --- | --- |
+| Install `shunt@portal` / `@spotify/portal-cli` | `npx thumbgate token-shunt-honesty --json` |
+| Buy Spotify Portal / Contact Sales as "make money" | Local slice (`offset`/`limit`, `cat \| grep`) |
+| Claim 90% token savings from App 162279530 | Measure `--lines` + `--returned-lines` |
+| Delegate architecture to Flash/Portal/AiKA | Frontier for reasoning; `local_slice` for boilerplate |
+| Spam buy links via All-repositories write | ECI: no ThumbGate paid outreach through Portal |
+
+## Reference
+
+- https://github.com/spotify/portal-ai-plugins (Apache-2.0 shunt plugin — we do not vendor it)
+- GitHub App install 162279530 = catalog connector, not revenue
+- Marketplace (free cite only): https://github.com/marketplace/actions/thumbgate-agent-governance
+
+```bash
+npx thumbgate token-shunt-honesty --json --lines=800
+npx thumbgate token-shunt-honesty --json --lines=800 --targeted
+npx thumbgate token-shunt-honesty --json --clone-portal
+```
+
+## Examples (show, don't tell)
+
+Weak: "Install Portal, we will save 90% and make money."
+
+Gold:
+
+```bash
+$ npx thumbgate token-shunt-honesty --json --lines=800
+{"ok":false,"findings":[{"id":"untargeted_bulk_read"}]}
+$ npx thumbgate token-shunt-honesty --json --lines=800 --targeted
+{"ok":true}
+```
+
+## Procedures
+
+1. Block untargeted Read above 350 lines.
+2. Block bare `cat`/`head`/`tail`/`less`/`more` of large files; allow pipes to grep/rg.
+3. Block returning the whole file to the frontier model.
+4. Refuse Portal/AiKA/Flash for architecture/debug.
+5. Treat GitHub App all-repos write as a warn, not a payout.
+
+## Rubric
+
+- `--lines=800` → `untargeted_bulk_read`
+- `--targeted` → `ok=true`
+- `--clone-portal` → `portal_clone_refused`
+- doctor: `npm run test:token-shunt-honesty` PASS
+- evidence: command output in the same turn

package/.agents/skills/typesafe-typed-questions-not-clone/SKILL.md

@@ -0,0 +1,82 @@
+---
+name: typesafe-typed-questions-not-clone
+description: >
+  TypeSafe (console.typesafe.ai/hook, Jev System One) is a calibrated decision
+  model, not a ThumbGate clone. Steal typed noul/choice/score + code-owned
+  route() + confidence-as-second-axis onto existing PreToolUse rails. Never
+  install typesafe-sdk, call api.typesafe.ai, clone Jev, or unpark the LLM
+  adjudicator (#3690/#3687). Slash: /typesafe-typed-questions-not-clone.
+---
+
+# TypeSafe typed questions — compare, do not clone
+
+## Goal
+
+Produce fail-closed typed-question composition for whom: ThumbGate PreToolUse
+— so a tool-call state is scored with independent noul/choice/score questions
+and **code** routes pass|review|block. TypeSafe/Jev stays a neighbor FORMAT,
+not a SKU, not an LLM adjudicator.
+
+## Constraints
+
+| NEVER | ALWAYS |
+| --- | --- |
+| Install `typesafe-sdk` / call `api.typesafe.ai` | `npx thumbgate typesafe-typed-questions --json` |
+| Clone Jev / System One as the gate | Deterministic matchers answer the battery |
+| Unpark LLM adjudicator (#3690/#3687) | Code owns `route()`; model does not emit the verdict |
+| One free-form "should we allow this?" judge | Atomic noul per hazard + score for severity |
+| Dual-edit untracked `llm-adjudicator` theater | Map onto existing `gate-check` rails |
+
+HARD fail closed. REFUSE SKU clones. ECI: no net-new governance product.
+Do NOT install `typesafe-sdk`. Do NOT call `api.typesafe.ai` from PreToolUse.
+
+Complementary to trading `/typesafe-system-one-not-clone` (AGENT-651 claim gate). Do **not** dual-edit `IgorGanapolsky/trading` `scripts/typesafe_claim_gate.py`. This skill is ThumbGate PreToolUse only.
+
+## Reference
+
+- https://console.typesafe.ai/hook (signed-in intro / Replay intro)
+- https://docs.typesafe.ai/cookbooks/llm_guardrails.md
+- https://docs.typesafe.ai/patterns/confidence-routing.md
+- Playground "Support agent audit" (noul battery + choice + score over one state)
+- `scripts/typesafe-typed-questions.js`
+- `/high-roi-steal-and-finish` · `/eci-thumbgate-ip-wall`
+
+## Examples (show, don't tell)
+
+Weak: Summarize Jev and `npm install typesafe-sdk` as the new PreToolUse hook.
+
+Gold:
+
+```bash
+$ npx thumbgate typesafe-typed-questions --json --map-only
+codeOwnsRoute: true
+$ npx thumbgate typesafe-typed-questions --json --tool-name=Bash --command='git push --force origin main'
+{"route":"block","answers":{"destructive":{"noul":1,"source":"deterministic"}}}
+$ node --test tests/typesafe-typed-questions.test.js
+# --clone-jev / --use-typesafe-api / --llm-adjudicate → fail
+```
+
+## Procedures
+
+```bash
+npx thumbgate typesafe-typed-questions --json --map-only
+npx thumbgate typesafe-typed-questions --json --tool-name=Bash --command='git push --force origin main'
+npx thumbgate typesafe-typed-questions --json --clone-jev
+npm run test:typesafe-typed-questions
+```
+
+1. Put the PreToolUse payload in `state`.
+2. Ask independent typed questions (noul per hazard, choice for family, score for severity).
+3. Answer them with deterministic matchers — never Jev.
+4. `route()` in code: action threshold → block, review threshold → review, severity ≥ 2 promotes review to block.
+5. Refuse `--clone-jev`, `--use-typesafe-api`, `--llm-adjudicate`, and model-emitted verdicts.
+
+## Rubric
+
+- empty / ordinary Read → `route=pass`, `ok=true`
+- `git push --force` → `route=block`, `destructive.noul=1`
+- `git add -A` → `route=review` (warn-level noul=0.55)
+- `--clone-jev` / `--use-typesafe-api` / `--llm-adjudicate` → `ok=false`
+- playground-shaped noul without a matcher → `unevaluated_question` (do not call Jev)
+- doctor: `npm run test:typesafe-typed-questions` PASS
+- evidence: command output in the same turn

package/.agents/skills/zvec-grep-compare-not-clone/SKILL.md

@@ -0,0 +1,34 @@
+---
+name: zvec-grep-compare-not-clone
+description: >
+  zg (zvec-grep) is a local-first search layer (ripgrep + BM25 + vector), not a
+  ThumbGate clone. Steal the four-route FORMAT onto existing rails; never install
+  @zvec/zvec-grep. Slash: /zvec-grep-compare-not-clone.
+---
+
+# zvec-grep — compare, do not clone
+
+## When
+Qwen zg, zvec-grep, MarkTechPost zvec, unify ripgrep BM25 vector, local-first search layer.
+
+## Do
+```bash
+npx thumbgate workspace-search-route --query="<intent or symbol>" --json
+npx thumbgate workspace-search-route --query="<symbol>" --rg --execute
+```
+
+Map routes to existing rails (see `docs/agents/code-search.md`):
+- `--rg` → ripgrep / Grep
+- `--fts` → filesystem-search / BM25F
+- `--vector` → LanceDB (local); remote needs `THUMBGATE_ALLOW_REMOTE_EMBED=1`
+- `--hybrid` → pragmatic-hybrid-search RRF (default)
+- `--graph` → Graphify AST graph
+
+## Never
+- `npm install -g @zvec/zvec-grep` inside ThumbGate
+- Claim ThumbGate is zg / zvec
+- Dual-edit DIRTY lesson-graph PR #3650
+- Silent remote embeddings
+
+## Source
+https://github.com/zvec-ai/zvec-grep (Apache-2.0) — FORMAT only; not affiliated.

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "thumbgate",
   "description": "One 👎 becomes a hard rule the agent cannot bypass. Captures thumbs-down feedback, distills it into PreToolUse Pre-Action Checks, enforced across every future Claude Code session.",
-  "version": "1.35.0",
+  "version": "1.37.2",
   "author": {
     "name": "Igor Ganapolsky",
     "email": "ig5973700@gmail.com",

package/.well-known/llms.txt

@@ -62,6 +62,7 @@ npx thumbgate init --agent claude-code
 - Codex Plugin: https://thumbgate.ai/codex-plugin
 - Free local CLI: 2 captures/day, 10 total captures, up to 3 active auto-promoted prevention rules, and local Pre-Action Checks after install (recall and lesson search are Pro-only)
 - Pro: $19/mo or $149/yr — personal enforcement proof, local dashboard, check debugger, DPO export, synced lessons/rules across machines, and review-ready exports
+- Managed AI Agent Workflow Gate: one-time engagement covering one supported workflow — a 60-minute review, one configured local gate, its regression test, and written rollout/rollback proof within two business days. Current terms: https://thumbgate.ai/pricing
 - Enterprise: custom pricing, scoped after intake — shared hosted lesson database, org-wide dashboard, GCP/DFCX guardrails, VPC gating, and white-glove workflow hardening sprints
 
 ## Links

package/.well-known/mcp/server-card.json

@@ -1,6 +1,6 @@
 {
   "name": "thumbgate",
-  "version": "1.35.0",
+  "version": "1.37.2",
   "description": "ThumbGate — 👍👎 feedback that teaches your AI agent. Thumbs down a mistake, it never happens again.",
   "homepage": "https://thumbgate.ai",
   "transport": "stdio",

package/CONTRIBUTING.md

@@ -0,0 +1,95 @@
+# Contributing to ThumbGate
+
+Thank you for your interest in improving ThumbGate.
+
+This project follows the [Contributor Covenant Code of Conduct](CODE_OF_CONDUCT.md).
+
+## How to Contribute
+
+1. Open an issue to discuss significant changes before writing code.
+2. Fork the repository and create a feature branch.
+3. Follow the existing two-space indentation and single-quote style.
+4. Add or update tests for any changed behavior.
+5. Run the relevant test suites (`npm test` and targeted tests) before opening a
+   pull request.
+6. Use [Conventional Commits](https://www.conventionalcommits.org/) for commit
+   messages.
+7. Reference the relevant issue in your PR description.
+
+## Contributor License Agreement / Developer Certificate of Origin
+
+By contributing code, documentation, diagrams, or other materials to the
+ThumbGate repository, you agree that:
+
+- You have the right to submit the contribution.
+- You grant ThumbGate and its maintainers a perpetual, worldwide, royalty-free,
+  irrevocable license to use, reproduce, modify, distribute, and sublicense your
+  contribution under the MIT License and any future dual-licensing arrangements
+  used by the project.
+- Your contribution is your original work or you have obtained all necessary
+  rights and permissions from any third-party rights holders.
+- You sign off each commit with `git commit -s`, certifying the
+  [Developer Certificate of Origin](https://developercertificate.org/):
+
+```text
+Developer Certificate of Origin
+Version 1.1
+
+Copyright (C) 2004, 2006 The Linux Foundation and its contributors.
+
+Everyone is permitted to copy and distribute verbatim copies of this
+license document, but changing it is not allowed.
+
+Developer's Certificate of Origin 1.1
+
+By making a contribution to this project, I certify that:
+
+(a) The contribution was created in whole or in part by me and I
+    have the right to submit it under the open source license
+    indicated in the file; or
+
+(b) The contribution is based upon previous work that, to the best
+    of my knowledge, is covered under an appropriate open source
+    license and I have the right under that license to submit that
+    work with modifications, whether created in whole or in part
+    by me, under the same open source license (unless I am
+    permitted to submit under a different license), as indicated
+    in the file; or
+
+(c) The contribution was provided directly to me by some other
+    person who certified (a), (b) or (c) and I have not modified
+    it.
+
+(d) I understand and agree that this project and the contribution
+    are public and that a record of the contribution (including all
+    personal information I submit with it, including my sign-off) is
+    maintained indefinitely and may be redistributed consistent with
+    this project or the open source license(s) involved.
+```
+
+Pull requests without a Signed-off-by line may be rejected.
+
+## Compensation and Commercial Use
+
+Contributions to the public repository are made on a volunteer basis unless a
+separate consulting agreement is signed with ThumbGate. The project maintainers
+may incorporate your contribution into both open-source and commercial hosted
+offerings governed by separate terms.
+
+## Employment / Confidentiality Boundary
+
+If you are subject to an employment agreement, non-disclosure agreement, or
+invention-assignment obligation that may overlap with ThumbGate's domain (for
+example, work in agentic AI, engineering acceleration, or AI governance), you
+must certify that each contribution is entirely your own work, created on your
+own time, without using any employer or third-party confidential information,
+trade secrets, equipment, or facilities. See
+`docs/legal/ECI_EMPLOYMENT_BOUNDARY_POLICY.md` for the project-wide policy.
+
+Your Signed-off-by line implicitly certifies this boundary in addition to the
+Developer Certificate of Origin.
+
+## Questions
+
+Contact the maintainers via the issue tracker at:
+https://github.com/IgorGanapolsky/ThumbGate/issues