diff --git a/AGENTS.md b/AGENTS.md
index 147433c..089678f 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -1,5 +1,9 @@
 ## Agent skills
 
+### Public Elixis skill
+
+Use `skills/elixis/SKILL.md` as the public repo skill for Source Corpus ingestion, pattern synthesis, identity, brand voice, design direction, naming research, and Market Kit workflows. Invoke `$elixis` in compatible agent hosts and keep the skill aligned with the CLI, MCP tools, HTTP API, and `CONTEXT.md` domain language.
+
 ### Issue tracker
 
 Issues tracked in GitHub (KyaniteLabs/Elixis) via `gh` CLI. See `agent-docs/issue-tracker.md`.
diff --git a/README.md b/README.md
index 620276b..3223de9 100644
--- a/README.md
+++ b/README.md
@@ -53,6 +53,10 @@ curl -X POST http://localhost:3110/api/ingest \
 
 Use the MCP tools `ingest_source` and `create_market_kit` for AI assistant integration.
 
+## Agent Skill
+
+Elixis includes a public agent skill at [`skills/elixis/SKILL.md`](skills/elixis/SKILL.md). Use `$elixis` in compatible agent hosts when you want an agent to choose the right CLI, HTTP API, or MCP workflow for Source Corpus ingestion, pattern synthesis, identity, brand voice, design direction, naming research, and Market Kit outputs.
+
 ## Configuration
 
 - `ELIXIS_INFERENCE_PROVIDER`: Inference provider (glm, kimi, openai)
@@ -61,6 +65,7 @@ Use the MCP tools `ingest_source` and `create_market_kit` for AI assistant integ
 
 ## Documentation
 
+- [skills/elixis/SKILL.md](skills/elixis/SKILL.md) — Public agent skill for Elixis workflows
 - [CONTEXT.md](CONTEXT.md) — Domain terminology and relationships
 - [docs/DEPLOY_SETUP.md](docs/DEPLOY_SETUP.md) — CI/CD deployment configuration
 - [agent-docs/](agent-docs/) — Agent and operator documentation
diff --git a/docs/llms.txt b/docs/llms.txt
index 3245de8..c181b81 100644
--- a/docs/llms.txt
+++ b/docs/llms.txt
@@ -21,5 +21,6 @@ Technical operator surfaces:
 - CLI: `python -m elixis run --text "..." --lens brand`
 - HTTP API: `python -m elixis serve --port 3110`; contract in `openapi.yaml`
 - MCP stdio adapter: `python -m elixis mcp`
+- Public agent skill: `skills/elixis/SKILL.md`, invoked as `$elixis` in compatible agent hosts
 
 Keywords: AI pattern synthesis, Source Corpus ingestion, Market Kit generator, AI identity generator, brand voice generator, design system generator, naming research tool, repository analysis for branding, marketing kit, SOUL.md generator, OpenClaw companion, local LLM tool.
diff --git a/llms.txt b/llms.txt
index 21504af..db8a207 100644
--- a/llms.txt
+++ b/llms.txt
@@ -8,5 +8,8 @@ A Python engine + CLI/HTTP/MCP surfaces that synthesizes patterns (identity, bra
 ## Install / use
 git clone the repo, then `pip install -e .` (not yet on PyPI).
 
+## Agent skill
+Invoke `$elixis` in compatible agent hosts when the user needs Source Corpus ingestion, pattern synthesis, identity, brand voice, design-system direction, naming research, or Market Kit outputs. The public skill lives at `skills/elixis/SKILL.md`.
+
 ## Links
 - Repository: https://github.com/KyaniteLabs/Elixis
diff --git a/skills/elixis/SKILL.md b/skills/elixis/SKILL.md
new file mode 100644
index 0000000..f25e052
--- /dev/null
+++ b/skills/elixis/SKILL.md
@@ -0,0 +1,70 @@
+---
+name: elixis
+description: Use Elixis for local-first pattern synthesis from source material into identity, brand voice, design-system direction, naming research, Source Corpus ingestion, and Market Kit outputs through CLI, HTTP API, or MCP tools.
+---
+
+# Elixis
+
+Use Elixis when an agent needs to turn scattered reference material, a GitHub repository, or a local folder into structured identity, brand, design, naming, or Market Kit outputs. Elixis builds a curated Source Corpus, resolves a pattern graph, and emits lens-specific results.
+
+## Start Here
+
+- Read `../../README.md` for installation, CLI, HTTP API, and MCP examples.
+- Read `../../CONTEXT.md` for domain language such as Source Target, Source Corpus, Corpus Signal, Market Kit, and Artifact Tier.
+- Read `../../llms.txt` for a compact public summary.
+- Use `../../elixis/mcp_server.py` when you need exact MCP tool names and schemas.
+- Use `../../elixis/__main__.py` when you need exact CLI command names and flags.
+
+## Choose A Surface
+
+- CLI: best for local operator work such as `ingest`, `run`, `extract`, `patterns`, `name`, `corpus inspect`, and `serve`.
+- MCP: best for agent hosts that need `ingest_source`, `create_market_kit`, `run_game`, `create_soul`, `extract_entities`, `analyze_patterns`, `research_name`, or `name_from_identity`.
+- HTTP API: best for local app/server workflows after `python -m elixis serve`.
+
+## Workflow
+
+1. Identify the input:
+   - raw reference text,
+   - a local folder path,
+   - a public or private GitHub repository URL,
+   - an existing ingestion run ID.
+2. Choose the output lens:
+   - `identity` for persona, SOUL-style, or self-model documents,
+   - `brand` for positioning, voice, and messaging,
+   - `design` for visual system direction,
+   - Market Kit for bundled naming, positioning, brand, messaging, design, and evidence.
+3. For repository or folder work, run ingestion first and keep the distinction between Source Target, Source Corpus, Corpus Signal, and Artifact.
+4. Request Artifact Tiers only when needed. Use `markdown`, `html`, `css`, or `market-page` intentionally instead of assuming all artifacts should be rendered.
+5. Preserve provenance and review sensitive-candidate behavior before using outputs in public strategy or launch material.
+
+## CLI Examples
+
+```bash
+elixis ingest --github https://github.com/owner/repo
+elixis ingest --path ./my-project --kit --artifact markdown --artifact market-page
+elixis corpus inspect <run-id>
+elixis run --file notes.md --lens brand --json
+elixis patterns --file notes.md
+elixis name --name "Elixis" --context "AI pattern synthesis engine"
+elixis mcp
+```
+
+## MCP Setup
+
+```json
+{
+  "mcpServers": {
+    "elixis": {
+      "command": "python",
+      "args": ["-m", "elixis.mcp_server"]
+    }
+  }
+}
+```
+
+## Guardrails
+
+- Do not treat an entire repository as raw prompt text. Build a Source Corpus first.
+- Do not include hidden files, large files, issues, PRs, commits, visual analysis, or code evidence unless they improve the requested output and the user allowed that evidence class.
+- Do not silently render HTML/CSS artifacts for every run; artifact generation is an explicit operator choice.
+- Treat Elixis outputs as synthesized recommendations grounded in evidence, not as final trademark, legal, or market validation.
diff --git a/skills/elixis/agents/openai.yaml b/skills/elixis/agents/openai.yaml
new file mode 100644
index 0000000..cffd486
--- /dev/null
+++ b/skills/elixis/agents/openai.yaml
@@ -0,0 +1,7 @@
+interface:
+  display_name: "Elixis"
+  short_description: "Source-corpus pattern synthesis"
+  default_prompt: "Use $elixis to synthesize this source material into identity, brand, design, naming, or Market Kit outputs."
+
+policy:
+  allow_implicit_invocation: true