diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
index ea64cfec7..c08859d49 100644
--- a/.claude-plugin/marketplace.json
+++ b/.claude-plugin/marketplace.json
@@ -118,6 +118,12 @@
       "source": "./plugins/partner-built/spglobal",
       "description": "S&P Global - Financial data and analytics skills including company tearsheets, earnings previews, and transaction summaries"
     },
+    {
+      "name": "aether",
+      "displayName": "Aether",
+      "source": "./plugins/partner-built/aether",
+      "description": "Agent-native search over SEC filings (10-K/10-Q/8-K), earnings-call transcripts, and a 29-act EU financial-regulation corpus — as-filed and citable, never analyst-adjusted, plus a third-party data marketplace."
+    },
     {
       "name": "claude-for-msft-365-install",
       "displayName": "Claude for Microsoft 365 Install",
diff --git a/plugins/partner-built/aether/.claude-plugin/plugin.json b/plugins/partner-built/aether/.claude-plugin/plugin.json
new file mode 100644
index 000000000..2cf68a0ab
--- /dev/null
+++ b/plugins/partner-built/aether/.claude-plugin/plugin.json
@@ -0,0 +1,25 @@
+{
+  "name": "aether",
+  "displayName": "Aether",
+  "version": "0.1.1",
+  "description": "Agent-native search over SEC filings (10-K/10-Q/8-K), earnings-call transcripts, and a 29-act EU financial-regulation corpus — as-filed and citable, never analyst-adjusted. Includes a two-sided data marketplace for third-party research.",
+  "author": {
+    "name": "EvidInvest (Aether)",
+    "email": "hao.huang@ebdsweden.com"
+  },
+  "homepage": "https://aether.evidinvest.com",
+  "repository": "https://github.com/EvidInvest/aether-developer",
+  "license": "Apache-2.0",
+  "keywords": [
+    "sec",
+    "edgar",
+    "filings",
+    "10-K",
+    "earnings-transcripts",
+    "eu-regulation",
+    "mifid",
+    "mica",
+    "financial-search",
+    "agent-native"
+  ]
+}
diff --git a/plugins/partner-built/aether/.mcp.json b/plugins/partner-built/aether/.mcp.json
new file mode 100644
index 000000000..29ed83d82
--- /dev/null
+++ b/plugins/partner-built/aether/.mcp.json
@@ -0,0 +1,8 @@
+{
+  "mcpServers": {
+    "aether": {
+      "type": "http",
+      "url": "https://api.aether.evidinvest.com/mcp"
+    }
+  }
+}
diff --git a/plugins/partner-built/aether/CONNECTORS.md b/plugins/partner-built/aether/CONNECTORS.md
new file mode 100644
index 000000000..ccfedec54
--- /dev/null
+++ b/plugins/partner-built/aether/CONNECTORS.md
@@ -0,0 +1,95 @@
+# Connectors
+
+This plugin connects to a single MCP server, **Aether** (`@evidinvest/aether-mcp`), which
+proxies to `https://api.aether.evidinvest.com`. No additional connectors are required. The
+server advertises **11 tools** in three families. Commands reference tools by their exact
+names below.
+
+## Tool Categories
+
+| Category | Tools | Description |
+|----------|-------|-------------|
+| Search | `financial_search`, `transcript_search`, `regulation_search` | First-party retrieval over SEC filings, earnings transcripts, and EU regulation |
+| Marketplace — read | `list_partners`, `partner_search`, `partner_proxy_search` | Discover and query third-party seller data |
+| Marketplace — sell | `seller_signup`, `seller_publish_document`, `seller_register_endpoint`, `seller_list_my_documents`, `seller_list_my_endpoints` | Publish documents or register paid endpoints as a seller |
+
+## Search Tools
+
+### `financial_search`
+Natural-language hybrid retrieval over SEC filings (10-K/10-Q/8-K) and supply-chain
+relationships. Returns ranked, structured, LLM-ready results with snippets, section titles,
+filing URLs, and tickers.
+
+| Param | Type | Notes |
+|-------|------|-------|
+| `query` | string | **Required.** Natural-language query. |
+| `domain` | enum | `public_equity` \| `supply_chain` \| `auto` (default `auto`). |
+| `limit` | number | Default 10, max 50. |
+| `fields` | string[] | Optional field projection. |
+| `profile` | enum | `bm25` \| `hybrid` \| `hybrid_rerank` \| `hybrid_rerank_tickerprior` (default — the production winner: hybrid + cross-encoder rerank + ticker-prior). |
+| `return_format` | enum | `section` (default — full SEC section the match belongs to) \| `chunk` (matching window only) \| `both`. |
+
+### `transcript_search`
+Hybrid retrieval over earnings-call transcripts (CEO/CFO commentary + analyst Q&A). Returns
+ranked verbatim speaker turns with citations. Use for forward-looking color (guidance,
+narrative) that filings don't capture.
+
+| Param | Type | Notes |
+|-------|------|-------|
+| `query` | string | **Required.** |
+| `ticker` | string | Optional ticker filter (e.g. `NVDA`). |
+| `lookback_quarters` | number | Keep calls within the last N quarters (default: no filter). |
+| `speaker_role` | string | Optional: `CEO` \| `CFO` \| `Analyst` \| `Operator`. |
+| `limit` | number | Default 10, max 50. |
+| `profile` | enum | `bm25` \| `hybrid` (default). |
+
+### `regulation_search`
+Search a 29-act EU financial-regulation corpus (~13,000 citable chunks). Returns ranked,
+citable units — Article-paragraphs, recitals, annex blocks — each with a human breadcrumb
+(e.g. `MiCA 2023/1114 · Art. 4 · para. 1`). Scoped to financial regulation, not all EU law.
+
+| Param | Type | Notes |
+|-------|------|-------|
+| `query` | string | **Required.** |
+| `celex` | string \| string[] | Optional CELEX filter, e.g. `32014L0065` (MiFID II), `32023R1114` (MiCA), `32013R0575` (CRR), `32022R2554` (DORA), `32024R1624` (AMLR). Omit to search the whole corpus. |
+| `doc_type` | string | `regulation` \| `directive` \| `rts` \| `its` \| `decision`. |
+| `article` | string | Single-article filter, e.g. `"20"` or `"12a"`. |
+| `chunk_type` | string | `paragraph` \| `article_intro` \| `recital` \| `table` \| `annex`. |
+| `aml_topics` | string \| string[] | `cdd` \| `edd` \| `pep` \| `str_reporting` \| `governance` \| `reporting` \| `transaction_monitoring`. |
+| `prefer_consolidated` | boolean | Prefer EUR-Lex consolidated text over original OJ text at equal relevance (hybrid only). |
+| `limit` | number | Default 10, max 50. |
+| `profile` | enum | `bm25` \| `hybrid` (default). |
+
+## Marketplace — Read Tools
+
+### `list_partners`
+List active marketplace sellers and what each offers. Each partner exposes zero or more modes:
+`indexed` (free queries via `partner_search`) and/or `proxy` (paid per-call queries via
+`partner_proxy_search`). Returns pricing so you can budget before calling. Params: `ticker`
+(optional coverage filter), `mode` (`indexed` \| `proxy` \| `any`), `limit`.
+
+### `partner_search`
+Search marketplace partner documents (research notes, supply-chain analyses) published by
+registered sellers. Returns ranked chunks with per-document attribution and license terms.
+Call `list_partners` first to scope via `partners`. Params: `query` (**required**), `partners[]`,
+`doc_types[]`, `ticker_filter`, `limit`, `profile`.
+
+### `partner_proxy_search`
+Route a query server-to-server to seller-registered API endpoints. **Costs accrue per call**
+regardless of result count; Aether holds seller credentials so the agent never sees the URL or
+token. Params: `query` (**required**), `partners[]` (**required** — you must pick explicitly,
+no surprise billing), `endpoints[]`, `ticker_filter`, `limit`, `confirm_charge` (must be `true`
+to execute; otherwise returns a dry-run quote).
+
+## Marketplace — Sell Tools
+
+For data publishers, not analysis workflows. Briefly:
+
+- **`seller_signup`** — create a seller account (returns an API key once; new accounts are
+  `pending_review` until ops approves, unless an `invite_code` is supplied).
+- **`seller_publish_document`** — publish/update a document (chunked, embedded, indexed in the
+  background); re-publishing the same `external_doc_id` replaces the prior version.
+- **`seller_register_endpoint`** — register a proxy endpoint (Mode B); secret stored encrypted
+  at rest, per-call pricing via `price_per_call_usd_cents`.
+- **`seller_list_my_documents`** / **`seller_list_my_endpoints`** — inventory + status; secrets
+  never returned.
diff --git a/plugins/partner-built/aether/README.md b/plugins/partner-built/aether/README.md
new file mode 100644
index 000000000..102a3aec2
--- /dev/null
+++ b/plugins/partner-built/aether/README.md
@@ -0,0 +1,82 @@
+# Aether — Agent-Native Financial Search
+
+Search SEC filings, earnings-call transcripts, and EU financial regulation in natural
+language, and get back **as-filed, citable** text — never analyst-adjusted. Aether is a
+financial-vertical search engine purpose-built for agents, exposed over MCP.
+
+## What This Plugin Does
+
+This plugin wires the [Aether MCP server](https://github.com/EvidInvest/aether-developer)
+into Claude and packages its tools into high-level workflows. Each command orchestrates a
+search tool into a complete, citation-backed answer instead of leaving you to call tools one
+at a time.
+
+The corpus today:
+
+- **SEC filings** — 10-K / 10-Q / 8-K for the S&P 500 across ~10 years, ~1.4M retrieval chunks.
+- **Earnings-call transcripts** — CEO / CFO commentary and analyst Q&A, for forward-looking
+  color that filings don't capture.
+- **EU financial regulation** — a 29-act corpus (~13,000 citable units) spanning MiFID II,
+  MiFIR, MAR, EMIR, MiCA, DORA, CRR/CRD IV, UCITS/AIFMD, PSD2, Solvency II, SFDR/Taxonomy/CSRD,
+  the 2024 AML package, and more.
+- **Marketplace** — third-party research and data published by registered sellers, searchable
+  alongside the first-party corpus.
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `/sec-deep-dive` | Pull the most relevant 10-K/10-Q/8-K sections on a topic for a company, with as-filed quotes and filing citations |
+| `/earnings-color` | Surface CEO/CFO/analyst commentary on a topic from earnings-call transcripts, quoted verbatim with speaker + date |
+| `/regulation-lookup` | Find the controlling EU-regulation text for a question, returned as citable Article-paragraphs / recitals |
+| `/supply-chain-exposure` | Find public companies disclosing exposure to a supply-chain risk, grouped by ticker |
+
+## Skills
+
+| Skill | Domain Knowledge |
+|-------|-----------------|
+| `sec-filings-research` | Retrieval profiles, `section` vs `chunk` returns, as-filed discipline, citation hygiene |
+| `earnings-transcript-research` | Speaker-role and lookback filtering, separating guidance from reported results |
+| `eu-regulation-research` | CELEX map, consolidated vs OJ text, doc-type/article/AML-topic filters, citation breadcrumbs |
+
+## Integrations
+
+This plugin connects to the **Aether MCP Server**, which serves these tool families:
+
+- **Search** — `financial_search` (SEC filings + supply-chain relationships),
+  `transcript_search` (earnings calls), `regulation_search` (EU financial regulation).
+- **Marketplace (read)** — `list_partners`, `partner_search` (indexed seller documents),
+  `partner_proxy_search` (paid server-to-server seller queries).
+- **Marketplace (sell)** — `seller_signup`, `seller_publish_document`,
+  `seller_register_endpoint`, `seller_list_my_documents`, `seller_list_my_endpoints`.
+
+See [CONNECTORS.md](CONNECTORS.md) for the complete tool reference and parameters.
+
+## Installation
+
+```
+claude plugins add aether
+```
+
+## Requirements & Authentication
+
+This is a **remote MCP server** over Streamable HTTP at
+`https://api.aether.evidinvest.com/mcp` (see [`.mcp.json`](.mcp.json)) — works in both
+Claude Cowork and Claude Code, nothing to install. Authentication:
+
+- **OAuth device code** — on first connect the client walks you through sign-in at
+  <https://aether.evidinvest.com>; the token is cached and refreshed automatically.
+  Verified accounts include a **3-month free trial** (5,000 calls/hr on Aether's first-party tools).
+- **Anonymous** — connect without signing in to try it out, rate-limited to 5 calls/hour.
+
+A self-hosted / stdio option (`npx @evidinvest/aether-mcp` with `AETHER_API_KEY`) is also
+available — see the [aether-developer repo](https://github.com/EvidInvest/aether-developer).
+
+## Why Aether
+
+- **As-filed, never adjusted.** Numbers and language come straight from the source filing, with
+  the section, filing URL, and ticker attached — not a vendor's normalized model.
+- **First agent-native financial search.** Built for tool-calling agents, not a human UI bolted
+  onto an API.
+- **One server, three corpora.** US disclosure, earnings narrative, and EU regulation behind a
+  single MCP connector, plus an open marketplace for third-party data.
diff --git a/plugins/partner-built/aether/commands/earnings-color.md b/plugins/partner-built/aether/commands/earnings-color.md
new file mode 100644
index 000000000..c83f1a75b
--- /dev/null
+++ b/plugins/partner-built/aether/commands/earnings-color.md
@@ -0,0 +1,42 @@
+---
+description: Surface CEO/CFO/analyst commentary on a topic from earnings-call transcripts, quoted verbatim with speaker and date
+argument-hint: "<topic> [ticker e.g. NVDA] [lookback quarters e.g. 4]"
+---
+
+# Earnings Color
+
+> Uses the Aether `transcript_search` tool. See [CONNECTORS.md](../CONNECTORS.md) for parameters.
+
+Pull forward-looking commentary — guidance, narrative, tone — from earnings-call transcripts
+that 10-K/10-Q filings don't capture. See the **earnings-transcript-research** skill for guidance
+on separating reported results from outlook.
+
+## Workflow
+
+### 1. Gather Input
+- Topic / question (required).
+- Optional: ticker, lookback window (quarters), speaker role of interest (CEO / CFO / Analyst).
+
+### 2. Search
+Call `transcript_search` with:
+- `query`: the topic phrased naturally.
+- `ticker`: if the user named a company.
+- `lookback_quarters`: if the user wants recent calls only (e.g. 4).
+- `speaker_role`: if they want only management (`CEO`/`CFO`) or only `Analyst` questions.
+- `limit`: 10.
+
+### 3. Synthesize
+- Quote speaker turns **verbatim**.
+- Attribute each: ticker, speaker name + role, and call date.
+- Distinguish management assertions from analyst questions; flag hedged or conditional language.
+
+## Output Format
+
+Lead with a 1-2 sentence read on management's stance, then:
+
+| Quote | Speaker (role) | Ticker | Call date |
+|-------|----------------|--------|-----------|
+| "…" | … (CEO) | … | … |
+
+Close by noting where commentary is consistent vs shifting across quarters, and any topics
+management notably avoided.
diff --git a/plugins/partner-built/aether/commands/regulation-lookup.md b/plugins/partner-built/aether/commands/regulation-lookup.md
new file mode 100644
index 000000000..d45000172
--- /dev/null
+++ b/plugins/partner-built/aether/commands/regulation-lookup.md
@@ -0,0 +1,44 @@
+---
+description: Find the controlling EU-regulation text for a question, returned as citable Article-paragraphs and recitals
+argument-hint: "<question> [act e.g. MiCA / MiFID II] [article e.g. 20]"
+---
+
+# Regulation Lookup
+
+> Uses the Aether `regulation_search` tool. See [CONNECTORS.md](../CONNECTORS.md) for parameters.
+
+Find the controlling text in the EU financial-regulation corpus and return it as **citable**
+units with breadcrumbs (e.g. `MiCA 2023/1114 · Art. 4 · para. 1`). See the
+**eu-regulation-research** skill for the CELEX map and consolidated-vs-OJ guidance.
+
+## Workflow
+
+### 1. Gather Input
+- Question (required).
+- Optional: which act (map to CELEX), specific article, doc type, AML topic.
+
+### 2. Search
+Call `regulation_search` with:
+- `query`: the question phrased naturally.
+- `celex`: if the user named an act — e.g. `32023R1114` (MiCA), `32014L0065` (MiFID II),
+  `32013R0575` (CRR), `32022R2554` (DORA), `32024R1624` (AMLR). See the skill for the full map.
+- `article`: if they cited a specific article.
+- `aml_topics`: for AML questions (e.g. `cdd`, `edd`, `pep`, `str_reporting`).
+- `prefer_consolidated`: `true` when the user wants current in-force text rather than as-published.
+- `limit`: 10.
+
+### 3. Synthesize
+- Quote the operative provision **verbatim**.
+- Always include the breadcrumb citation returned with each unit.
+- If consolidated and original text differ materially, note both.
+
+## Output Format
+
+Lead with a 1-2 sentence answer grounded in the provision, then:
+
+| Provision (verbatim) | Citation | Type |
+|----------------------|----------|------|
+| "…" | MiCA 2023/1114 · Art. 4 · para. 1 | paragraph |
+
+Close with a plain-language reading. Flag that this is the corpus text, not legal advice, and
+note if the question spans acts not in the 29-act corpus.
diff --git a/plugins/partner-built/aether/commands/sec-deep-dive.md b/plugins/partner-built/aether/commands/sec-deep-dive.md
new file mode 100644
index 000000000..7efecd76c
--- /dev/null
+++ b/plugins/partner-built/aether/commands/sec-deep-dive.md
@@ -0,0 +1,45 @@
+---
+description: Pull the most relevant 10-K/10-Q/8-K sections on a topic for a company, with as-filed quotes and filing citations
+argument-hint: "<ticker e.g. AAPL> <topic e.g. supply-chain risk in Taiwan>"
+---
+
+# SEC Deep Dive
+
+> Uses the Aether `financial_search` tool. See [CONNECTORS.md](../CONNECTORS.md) for parameters.
+
+Search a company's SEC filings for a topic and return the most relevant **as-filed** sections,
+quoted verbatim with citations. See the **sec-filings-research** skill for retrieval-profile and
+return-format guidance.
+
+## Workflow
+
+### 1. Gather Input
+- Ticker (required).
+- Topic / question (required).
+- Optional: filing-type focus (10-K vs 10-Q vs 8-K), recency preference.
+
+### 2. Search
+Call `financial_search` with:
+- `query`: the topic phrased naturally, scoped to the company (e.g. "Apple supply-chain concentration risk in Taiwan").
+- `domain`: `public_equity`.
+- `return_format`: `section` (full section context for accurate quoting).
+- `profile`: leave default (`hybrid_rerank_tickerprior`).
+- `limit`: 10.
+
+If results look thin, widen the query or retry with `return_format: "both"`.
+
+### 3. Synthesize
+- Quote the relevant passages **verbatim** — never paraphrase numbers or risk language.
+- Attribute every quote: ticker, filing type, section title, and filing URL from the result.
+- Group by theme; note disagreements or changes in language across filings if visible.
+
+## Output Format
+
+Lead with a 1-2 sentence answer to the question, then:
+
+| Quote (as-filed) | Filing | Section | Date / URL |
+|------------------|--------|---------|------------|
+| "…" | 10-K | Risk Factors | … |
+
+Close with a short synthesis: what the filings actually say, and any gaps where the disclosure
+is silent. Do not infer beyond the cited text.
diff --git a/plugins/partner-built/aether/commands/supply-chain-exposure.md b/plugins/partner-built/aether/commands/supply-chain-exposure.md
new file mode 100644
index 000000000..31fdd7262
--- /dev/null
+++ b/plugins/partner-built/aether/commands/supply-chain-exposure.md
@@ -0,0 +1,38 @@
+---
+description: Find public companies disclosing exposure to a supply-chain risk, grouped by ticker
+argument-hint: "<risk e.g. rare-earth magnets from China>"
+---
+
+# Supply-Chain Exposure
+
+> Uses the Aether `financial_search` tool in supply-chain mode. See [CONNECTORS.md](../CONNECTORS.md).
+
+Find public companies whose SEC filings disclose exposure to a given supply-chain risk, then
+rank and group the evidence by ticker. See the **sec-filings-research** skill for retrieval
+guidance.
+
+## Workflow
+
+### 1. Gather Input
+- The risk or dependency (required) — e.g. a region, supplier, input, or chokepoint.
+
+### 2. Search
+Call `financial_search` with:
+- `query`: the risk phrased naturally (e.g. "dependence on rare-earth magnets sourced from China").
+- `domain`: `supply_chain`.
+- `return_format`: `section`.
+- `limit`: 25 (cast wider — you're surveying the universe, not one company).
+
+### 3. Synthesize
+- Group hits by ticker; keep the single most relevant **as-filed** passage per company.
+- Rank by relevance and recency.
+- Quote verbatim with filing citation; do not infer exposure that isn't disclosed.
+
+## Output Format
+
+| Ticker | Most relevant disclosure (as-filed) | Filing | Date |
+|--------|--------------------------------------|--------|------|
+| … | "…" | 10-K | … |
+
+Close with a short read on where exposure clusters, and the limits of the search (only companies
+that *disclose* the risk in the covered corpus appear).
diff --git a/plugins/partner-built/aether/skills/earnings-transcript-research/SKILL.md b/plugins/partner-built/aether/skills/earnings-transcript-research/SKILL.md
new file mode 100644
index 000000000..410524ec6
--- /dev/null
+++ b/plugins/partner-built/aether/skills/earnings-transcript-research/SKILL.md
@@ -0,0 +1,44 @@
+---
+name: earnings-transcript-research
+description: Search earnings-call transcripts with Aether's transcript_search tool for forward-looking management commentary and analyst Q&A. Use when you need guidance, narrative, tone, or color that 10-K/10-Q filings don't capture, quoted verbatim with speaker and call date.
+---
+
+# Earnings Transcript Research with Aether
+
+You are an analyst mining earnings calls for what management *said* — guidance, framing, tone,
+and the questions analysts pressed on. This is the forward-looking layer that audited filings
+omit. Aether's `transcript_search` tool returns ranked verbatim speaker turns with citations.
+
+## Core Principle: Quote the Speaker, Date the Claim
+
+Management commentary is only useful if it's attributed and time-stamped. Quote speaker turns
+verbatim, label the speaker and role, and stamp the call date. Distinguish a CEO assertion from
+an analyst's premise — they carry very different weight.
+
+## The `transcript_search` Tool
+
+| Param | Guidance |
+|-------|----------|
+| `query` | Natural language; topic-first ("data-center demand durability"). |
+| `ticker` | Scope to a company (e.g. `NVDA`) when the question is single-name. |
+| `lookback_quarters` | Restrict to recent calls (e.g. 4) when recency matters; omit for full history. |
+| `speaker_role` | `CEO` / `CFO` for management view; `Analyst` to see what the sell-side is probing; `Operator` rarely useful. |
+| `limit` | 10 default; raise for cross-company theme sweeps. |
+| `profile` | `hybrid` (default) for topical questions; `bm25` for exact-phrase. |
+
+## Workflow
+
+1. **Frame** — guidance vs. reported results vs. analyst concern? Choose `speaker_role` accordingly.
+2. **Scope time** — set `lookback_quarters` if the user cares about the latest stance, not history.
+3. **Retrieve** — call `transcript_search`.
+4. **Separate signal** — split management claims from analyst questions; flag hedged/conditional
+   language ("we expect", "should", "if conditions hold").
+5. **Cite** — speaker name + role + ticker + call date for every quote.
+
+## Anti-Patterns
+
+- Presenting an analyst's framing as management's view.
+- Dropping the call date — outlook commentary ages fast.
+- Smoothing hedged language into a hard commitment.
+- Mixing transcript color with as-filed numbers without labeling which is which (use the
+  `sec-filings-research` skill for the filed figures).
diff --git a/plugins/partner-built/aether/skills/eu-regulation-research/SKILL.md b/plugins/partner-built/aether/skills/eu-regulation-research/SKILL.md
new file mode 100644
index 000000000..4e4d26e52
--- /dev/null
+++ b/plugins/partner-built/aether/skills/eu-regulation-research/SKILL.md
@@ -0,0 +1,74 @@
+---
+name: eu-regulation-research
+description: Search EU financial regulation with Aether's regulation_search tool across a 29-act corpus (MiFID II, MiCA, DORA, CRR, AMLR, and more) and return citable Article-paragraphs, recitals, and annex blocks. Use for compliance questions grounded in the actual regulatory text rather than summaries.
+---
+
+# EU Regulation Research with Aether
+
+You are a regulatory analyst answering from the **text of the regulation**, not from memory or
+summary. Aether's `regulation_search` tool covers a 29-act EU financial-regulation corpus
+(~13,000 citable units) and returns ranked Article-paragraphs, recitals, and annex blocks, each
+with a human breadcrumb citation (e.g. `MiCA 2023/1114 · Art. 4 · para. 1`).
+
+## Core Principle: Cite the Provision, Note the Version
+
+Quote the operative provision verbatim and always include the breadcrumb the tool returns.
+Regulation has versions: the as-published Official Journal text and later EUR-Lex consolidated
+text can differ. Be explicit about which you're quoting.
+
+## CELEX Quick Map
+
+Use `celex` to pin a search to one act:
+
+| Act | CELEX |
+|-----|-------|
+| MiFID II | `32014L0065` |
+| MiFIR | `32014R0600` |
+| MAR (Market Abuse) | `32014R0596` |
+| EMIR | `32012R0648` |
+| MiCA | `32023R1114` |
+| DORA | `32022R2554` |
+| CRR | `32013R0575` |
+| CRD IV | `32013L0036` |
+| UCITS | `32009L0065` |
+| AIFMD | `32011L0061` |
+| PSD2 | `32015L2366` |
+| Solvency II | `32009L0138` |
+| SFDR | `32019R2088` |
+| AMLR (2024) | `32024R1624` |
+| GDPR | `32016R0679` |
+
+The corpus also covers Prospectus, CSDR, Short-Selling, SFTR, Benchmarks, BRRD, SRMR, IDD,
+Taxonomy, CSRD, the rest of the 2024 AML package (AMLAR, AMLD6, Transfer-of-Funds), and Credit
+Rating Agencies. It is scoped to financial regulation — not all EU law.
+
+## The `regulation_search` Tool
+
+| Param | Guidance |
+|-------|----------|
+| `query` | Natural-language question. |
+| `celex` | Pin to one or more acts (see map). Omit to search the whole corpus. |
+| `doc_type` | `regulation` / `directive` / `rts` / `its` / `decision`. |
+| `article` | Narrow to a single article, e.g. `"20"` or `"12a"`. |
+| `chunk_type` | `paragraph` / `article_intro` / `recital` / `table` / `annex`. |
+| `aml_topics` | For AML questions: `cdd`, `edd`, `pep`, `str_reporting`, `governance`, `reporting`, `transaction_monitoring`. |
+| `prefer_consolidated` | `true` to favour in-force consolidated text over original OJ text (hybrid only). |
+| `profile` | `hybrid` (default) for topical questions; `bm25` for exact-phrase. |
+
+## Workflow
+
+1. **Localize** — name the likely act(s) and set `celex`; this sharply improves precision.
+2. **Choose the version** — `prefer_consolidated: true` for "what's in force now"; leave off for
+   as-published text.
+3. **Retrieve** — call `regulation_search`; narrow with `article`/`chunk_type` if the answer
+   should be a specific provision (operative `paragraph`) vs. context (`recital`).
+4. **Quote + cite** — verbatim provision plus its breadcrumb.
+5. **Read plainly** — give a plain-language reading, but anchor it to the cited text.
+
+## Anti-Patterns
+
+- Answering from general knowledge instead of retrieved text.
+- Dropping the breadcrumb citation.
+- Conflating a `recital` (interpretive context) with an operative `paragraph` (the binding rule).
+- Implying the answer is legal advice — it's corpus text. Flag when a question reaches beyond the
+  29-act corpus.
diff --git a/plugins/partner-built/aether/skills/sec-filings-research/SKILL.md b/plugins/partner-built/aether/skills/sec-filings-research/SKILL.md
new file mode 100644
index 000000000..381983041
--- /dev/null
+++ b/plugins/partner-built/aether/skills/sec-filings-research/SKILL.md
@@ -0,0 +1,45 @@
+---
+name: sec-filings-research
+description: Search SEC filings (10-K/10-Q/8-K) and supply-chain relationships with Aether's financial_search tool and return as-filed, citable evidence. Use when researching a company's disclosures, risk factors, MD&A, segment detail, or supply-chain exposure from primary filings rather than analyst-adjusted data.
+---
+
+# SEC Filings Research with Aether
+
+You are a disclosure-research analyst. Your job is to retrieve what companies *actually filed*
+and present it with citations — never to substitute a normalized or analyst-adjusted number for
+the as-filed text. Aether's `financial_search` tool serves SEC filings (10-K/10-Q/8-K) and
+supply-chain relationships over a corpus of ~1.4M chunks.
+
+## Core Principle: As-Filed, Always Cited
+
+Every claim must trace to a filing. Quote the operative language verbatim, attach the section
+title and filing URL the tool returns, and resist the urge to round, restate, or "clean up"
+reported figures. If the filing is silent on something, say so — do not infer.
+
+## The `financial_search` Tool
+
+| Param | Guidance |
+|-------|----------|
+| `query` | Phrase naturally and scope to the company when you can ("Nvidia data-center segment concentration"). |
+| `domain` | `public_equity` for filings; `supply_chain` for relationship/exposure queries; `auto` if unsure. |
+| `profile` | Leave default (`hybrid_rerank_tickerprior`) — hybrid retrieval + cross-encoder rerank + ticker-prior boost is the production winner. Drop to `bm25` only for exact-phrase lookups. |
+| `return_format` | `section` (default) for quoting in context; `chunk` for a tight matching window; `both` when you need the match *and* its surrounding section. |
+| `limit` | 10 for a single-company question; 25+ when surveying many companies (e.g. supply-chain sweeps). |
+| `fields` | Project specific fields only when you need a compact response. |
+
+## Workflow
+
+1. **Scope** — single company or a universe? Single → tight query + `limit` 10. Universe →
+   `domain: supply_chain`, broader query, higher `limit`.
+2. **Retrieve** — call `financial_search`; prefer `return_format: "section"` so quotes keep context.
+3. **Verify** — confirm each passage actually addresses the question before quoting it. Discard
+   loosely-related hits rather than padding the answer.
+4. **Cite** — ticker + filing type + section + URL/date for every quote.
+5. **Synthesize** — lead with a direct answer, support with verbatim quotes, and flag silences.
+
+## Anti-Patterns
+
+- Paraphrasing reported numbers (breaks the as-filed guarantee).
+- Quoting a `chunk` without checking the surrounding `section` for contradicting context.
+- Inferring exposure or intent the filing doesn't state.
+- Treating absence of a hit as proof of absence — note corpus limits (S&P 500, ~10 years).