@heretek-ai/epistemic-swarm 0.2.3 → 0.3.0

package/.agents/skills/brainstorming/SKILL.md

@@ -0,0 +1,13 @@
+# Brainstorming (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/brainstorming/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `iumbtems_brainstorm`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.agents/skills/code_audit/SKILL.md

@@ -0,0 +1,13 @@
+# Code Audit (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/code_audit/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `iumbtems_code_audit`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.agents/skills/epistemic_search/SKILL.md

@@ -0,0 +1,13 @@
+# Epistemic Search (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/epistemic_search/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `brave_web_search / firecrawl_scrape (research MCP servers)`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.agents/skills/grilling/SKILL.md

@@ -0,0 +1,13 @@
+# Socratic Grilling (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/grilling/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `iumbtems_socratic_frontier`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.agents/skills/oss_scout/SKILL.md

@@ -0,0 +1,13 @@
+# OSS Scout (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/oss_scout/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `iumbtems_oss_scout`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.agents/skills/research_cache/SKILL.md

@@ -0,0 +1,13 @@
+# Research Cache (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/research_cache/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `iumbtems_verify_quote`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.agents/skills/swarm_config/SKILL.md

@@ -0,0 +1,13 @@
+# Swarm Config (thin adapter stub)
+
+This file is a POINTER, not the implementation. It exists so harness skill
+discovery finds an entry; the real skill lives in the IUMBTEMS repo.
+
+- Canonical prose & scripts: `skills/swarm_config/`
+- Canonical programmatic surface: `python3 runner/mcp_server.py` (stdio MCP),
+  or one-shot: `python3 runner/mcp_server.py call <tool> '{...json...}'`
+- MCP tools for this skill: `iumbtems_config`
+
+Epistemic rules apply regardless of harness: tag claims as
+`[VERIFIED: <hash>]`, `[INFERRED: <reasoning>]`, `[HYPOTHESIS: <test>]`, or
+`[NEGATIVE_KNOWLEDGE: <query>]`. Writes go only to `.research/`.

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "epistemic-swarm",
-  "version": "0.2.3",
+  "version": "0.3.0",
   "description": "High-Integrity Dialectic Research Agent Harness for Claude Code enforcing empirical evidence over parametric hallucination.",
   "author": {
     "name": "Heretek AI",
@@ -42,15 +42,24 @@
   "mcpServers": {
     "brave-search": {
       "command": "python3",
-      "args": ["${CLAUDE_PLUGIN_ROOT}/config/mcp_launcher.py", "brave-search"]
+      "args": [
+        "${CLAUDE_PLUGIN_ROOT}/config/mcp_launcher.py",
+        "brave-search"
+      ]
     },
     "firecrawl": {
       "command": "python3",
-      "args": ["${CLAUDE_PLUGIN_ROOT}/config/mcp_launcher.py", "firecrawl"]
+      "args": [
+        "${CLAUDE_PLUGIN_ROOT}/config/mcp_launcher.py",
+        "firecrawl"
+      ]
     },
     "searxng": {
       "command": "python3",
-      "args": ["${CLAUDE_PLUGIN_ROOT}/config/mcp_launcher.py", "searxng"]
+      "args": [
+        "${CLAUDE_PLUGIN_ROOT}/config/mcp_launcher.py",
+        "searxng"
+      ]
     }
   },
   "skills": [
@@ -59,6 +68,7 @@
     "./skills/epistemic_search",
     "./skills/swarm_config",
     "./skills/code_audit",
-    "./skills/oss_scout"
+    "./skills/oss_scout",
+    "./skills/brainstorming"
   ]
 }

package/.omp/README.md

@@ -0,0 +1,39 @@
+# IUMBTEMS for OMP (oh-my-pi / `omp.sh`)
+
+Project-local OMP surface. Global install mirrors live under
+`~/.oh-omp/agent/{commands,prompts,hooks}/` (copy these files there for
+every-project availability).
+
+## Install as OMP plugin package
+
+This repo declares both manifest keys in `package.json`:
+
+```json
+{
+  "pi": { "extensions": ["./extensions/pi/index.js"], "skills": [...], "prompts": [...] },
+  "omp": { "extensions": ["./extensions/pi/index.js"], "skills": [...], "prompts": [...] }
+}
+```
+
+```bash
+omp install npm:@heretek-ai/epistemic-swarm
+# local dev:
+omp -e ./extensions/pi/index.js
+omp list
+tail -F ~/.omp/logs/omp.$(date +%F).log  # extension load errors land here, not the TUI
+```
+
+## Project-local surfaces in this directory
+
+```
+.omp/
+├── README.md            # this file
+├── SYSTEM.md            # project system-prompt override (safe subset of base_epistemic_system.md)
+├── commands/*.md        # reusable slash commands: /swarm /grill /audit /scout /brainstorming /swarm-config
+├── prompts/*.md         # prompt templates with `description:` frontmatter + $1/$@ expansion
+└── hooks/pre/*.ts       # pre-tool hooks (epistemic redirect)
+└── hooks/post/*.ts      # post-tool hooks (audit-trail log)
+```
+
+Commands return a string sent as the LLM prompt, or execute shell via the
+Bun worker and return output. All runners write only under `.research/`.

package/.omp/SYSTEM.md

@@ -0,0 +1,12 @@
+# IUMBTEMS project instructions (OMP SYSTEM.md override)
+
+You operate under the Epistemic Integrity Protocol (see `prompts/base_epistemic_system.md`):
+
+- Parametric memory is quarantined. Fetch, verify, then assert.
+- Tag every factual claim: `[VERIFIED:<hash>]`, `[INFERRED:<chain>]`, `[HYPOTHESIS:<falsification>]`, `[NEGATIVE_KNOWLEDGE:<query>]`.
+- Cache every source to `.research/sources/<sha256>.md` before citing
+  (`python3 skills/research-cache/hasher.py cache --url "<URL>" --content "$(cat file.md)"`).
+- Skills: `skills/{grilling,research_cache,epistemic_search,swarm_config,code_audit,oss_scout,brainstorming}/SKILL.md`.
+- Runners (Bun/Python kernels may call back into agent tools read/search/task):
+  `python3 runner/research_swarm.py --mode <research|audit|scout|hybrid|brainstorm> --objective "<...>"`.
+- `/brainstorming` MUST produce feature vectors + lateral moves + falsifiable spikes, never bug-fix lists.

package/.omp/commands/audit.md

@@ -0,0 +1,10 @@
+---
+description: Dialectic codebase architecture and security audit with line-level proof
+---
+
+Run the IUMBTEMS dialectic codebase audit on `$1`:
+
+`python3 runner/research_swarm.py --mode audit --objective "$1"`
+
+Report: `.research/code_audit_report.md` with `file:///path#Lstart-Lend` pointers
+and verbatim snippets. Every vulnerability needs a concrete patch blueprint.

package/.omp/commands/brainstorming.md

@@ -0,0 +1,12 @@
+---
+description: Lateral brainstorming — novel feature vectors and speculative spikes
+---
+
+Run IUMBTEMS lateral brainstorming on `$1` (see `skills/brainstorming/SKILL.md`
+and `prompts/agent_brainstormer.md`):
+
+1. Ingest workspace context read-only (`python3 skills/brainstorming/scripts/brainstorm.py --objective "$1" --show-context`).
+2. Thesis / Radical Antithesis / Synthesis — never bug-fix lists.
+3. Output Novel Feature Vectors (3-5), Lateral Architectural Moves (2-3),
+   Experimental Hypotheses (2-3 with `[HYPOTHESIS:<falsification test>]`).
+4. Write `.research/brainstorm_report.md` (or run `python3 runner/research_swarm.py --mode brainstorm --objective "$1"`).

package/.omp/commands/grill.md

@@ -0,0 +1,9 @@
+---
+description: Socratic grilling and decision frontier exploration
+---
+
+Launch IUMBTEMS Socratic grilling for `$1`.
+
+Follow `skills/grilling/SKILL.md` (rounds, frontier, assumption inversion,
+factual vs decisional separation). Persist with
+`python3 skills/grilling/socratic_tree.py --export`.

package/.omp/commands/scout.md

@@ -0,0 +1,10 @@
+---
+description: Scout open-source libraries and build clean-room blueprints
+---
+
+Scout open-source solutions for `$1`:
+
+`python3 runner/research_swarm.py --mode scout --objective "$1"`
+
+Flag GPL/AGPL contamination, solo-maintainer risk, dependency weight, CVEs.
+Report: `.research/oss_scout_report.md` plus a clean-room re-implementation blueprint.

package/.omp/commands/swarm-config.md

@@ -0,0 +1,10 @@
+---
+description: Inspect or tune Epistemic Swarm parameters (engine, depth, mode)
+---
+
+Inspect or tune IUMBTEMS swarm parameters: `$1`.
+
+`python3 skills/swarm_config/configure.py $1`
+
+Engines: `duckduckgo` (zero-key default), `brave`, `firecrawl`, `searxng`.
+Modes: `research`, `audit`, `scout`, `hybrid`, `brainstorm`. Depth: 1-4.

package/.omp/commands/swarm.md

@@ -0,0 +1,9 @@
+---
+description: Run IUMBTEMS dialectic research swarm on an objective
+---
+
+Run the IUMBTEMS dialectic research swarm on `$1`.
+
+1. If the objective is ambiguous, follow `skills/grilling/SKILL.md` first.
+2. Execute `python3 runner/research_swarm.py --objective "$1" $@[2:]`.
+3. Summarize `.research/final_synthesis.md`, preserving `[VERIFIED:<hash>]` pointers.

package/.omp/hooks/post/epistemic-audit.ts

@@ -0,0 +1,16 @@
+import type { HookAPI } from "@oh-my-pi/pi-coding-agent/hooks";
+
+/**
+ * OMP post-tool hook: remind the agent to keep the epistemic audit trail.
+ * Every [VERIFIED:<hash>] quote must exist verbatim in .research/sources/<sha256>.md.
+ */
+export async function postTool(api: HookAPI) {
+  api.on("tool:after", (event) => {
+    if (event?.ok === false) {
+      // Surface failures as negative knowledge, never parametric guesses.
+      event.note?.(
+        "Record tool failures as [NEGATIVE_KNOWLEDGE:<query>] with the exact query — do not hallucinate a substitute answer."
+      );
+    }
+  });
+}

package/.omp/hooks/pre/epistemic-redirect.ts

@@ -0,0 +1,21 @@
+import type { HookAPI } from "@oh-my-pi/pi-coding-agent/hooks";
+
+/**
+ * OMP pre-tool hook: redirect unverified web tools into the epistemic pipeline.
+ * Mirrors hooks/hooks.json (Claude Code) for the omp.sh hook layout.
+ */
+export async function preTool(api: HookAPI) {
+  api.on("tool:before", (event) => {
+    const name = String(event?.tool ?? "").toLowerCase();
+    if (name.includes("websearch") || name.includes("web_search")) {
+      event.redirect?.(
+        'Use epistemic search instead: python3 skills/epistemic_search/scripts/search.py "<query>" (SHA-256 cached, verifiable).'
+      );
+    }
+    if (name.includes("webfetch") || name.includes("web_fetch")) {
+      event.redirect?.(
+        'Use epistemic fetch instead: python3 skills/epistemic_search/scripts/fetch.py "<url>" (content-addressed into .research/sources/).'
+      );
+    }
+  });
+}

package/.omp/prompts/brainstorming.md

@@ -0,0 +1,8 @@
+---
+description: Lateral brainstorming prompt template — what-if features and spikes
+---
+
+Brainstorm laterally on `$1` per `skills/brainstorming/SKILL.md`:
+Novel Feature Vectors (3-5), Lateral Architectural Moves (2-3),
+Experimental Hypotheses (2-3 with falsification criteria).
+Context: `$@[2:]`. Never output bug-fix lists.

package/.omp/prompts/swarm.md

@@ -0,0 +1,8 @@
+---
+description: Run IUMBTEMS dialectic research swarm on an objective
+---
+
+Run the IUMBTEMS dialectic research swarm on `$1`.
+
+Execute `python3 runner/research_swarm.py --objective "$1" $@[2:]` and
+summarize `.research/final_synthesis.md` with `[VERIFIED:<hash>]` pointers.

package/MARKETPLACE.md

@@ -74,6 +74,14 @@ claude plugin marketplace list
 
 ---
 
+## 💡 Lateral Brainstorming (`/brainstorming`)
+
+Divergent ideation subagent (Thesis / Radical Antithesis / Synthesis) producing
+Novel Feature Vectors, Lateral Architectural Moves, and falsifiable spike hypotheses.
+Skill: `skills/brainstorming/SKILL.md` (`/brainstorming <prompt>`), system prompt:
+`prompts/agent_brainstormer.md`, runner: `python3 runner/research_swarm.py --mode brainstorm --objective "<prompt>" --mock-claude`.
+Available on all seven harnesses (Claude Code, OpenCode `iumbtems_brainstorm`, Pi/OMP `/brainstorming`, Gemini `/brainstorming`, Codex `$brainstorming`, AntiGravity skill).
+
 ## 🔍 Inspecting Installed Plugins
 
 To view installed plugins and their active components:

package/README.md

@@ -6,7 +6,7 @@
 [![Marketplace Validated](https://github.com/Heretek-AI/IUMBTEMS/actions/workflows/validate-marketplace.yml/badge.svg)](https://github.com/Heretek-AI/IUMBTEMS/actions)
 
 > **High-Integrity Dialectic Research Agent Harness**  
-> *Universal support for Claude Code, Pi (pi.dev), and OpenCode V2.*  
+> *Universal support for Claude Code, OpenCode V2, Pi (pi.dev), OMP (oh-my-pi), Gemini CLI, Codex CLI, and AntiGravity.*  
 > *Enforcing verified empirical evidence over parametric hallucination.*
 
 ---
@@ -26,7 +26,8 @@ Modern LLMs suffer from parametric hallucination, sycophancy, and premature narr
 
 ## 🌐 Universal Multi-Platform Support
 
-IUMBTEMS is packaged as a single universal npm package (`@heretek-ai/epistemic-swarm` with `iumbtems` binary) that runs across the three major autonomous agent platforms:
+IUMBTEMS is packaged as a single universal npm package (`@heretek-ai/epistemic-swarm` with `iumbtems` binary) that runs across seven major autonomous agent harnesses.
+Single source of truth: `skills/*` + `prompts/*`. The canonical programmatic surface is the first-party MCP server (`python3 runner/mcp_server.py`), which exposes the `iumbtems_*` tools in-process. Harness targets register that server and carry only thin skill stubs generated with `python3 scripts/build_adapters.py` (`iumbtems adapters`) into `plugins/*`, `.agents/skills` — stubs point at the canonical `skills/*` prose instead of copying code.
 
 ### 1. Claude Code (Native Marketplace & Overlay)
 
@@ -70,7 +71,8 @@ pi install npm:@heretek-ai/epistemic-swarm
   - `/swarm-config`: Inspect or tune search engine, depth (1-4), and mode
   - `/audit <target>`: Run dialectic codebase architecture and security audit
   - `/scout <feature>`: Scout open-source software and clean-room blueprints
-- **Agent Tools**: `iumbtems_config`, `iumbtems_verify_quote`.
+- **Agent Tools**: `iumbtems_config`, `iumbtems_verify_quote`, `iumbtems_brainstorm`.
+- OMP (`omp.sh`, oh-my-pi) shares the same entry point: `omp install npm:@heretek-ai/epistemic-swarm`, project commands in `.omp/commands/` (`/swarm`, `/grill`, `/audit`, `/scout`, `/brainstorming`, `/swarm-config`), prompts in `.omp/prompts/`, hooks in `.omp/hooks/pre|post/`.
 - All commands automatically respect `.research/config.json`.
 
 ### 3. OpenCode V2 (`opencode.ai`)
@@ -96,9 +98,33 @@ OpenCode V2 automatically registers the full tool suite:
 - `iumbtems_swarm_research`: Dispatches dialectic researcher pairs (Thesis vs. Antithesis).
 - `iumbtems_verify_quote`: Audits verbatim citations against the SHA-256 source cache.
 - `iumbtems_socratic_frontier`: Advances the Socratic decision tree frontier.
+- `iumbtems_brainstorm`: Lateral ideation portfolio (feature vectors + paradigm moves + falsifiable spikes).
 
 *(See [config/opencode-snippet.json](config/opencode-snippet.json) for custom agent definitions).*
 
+### 4. Gemini CLI
+```bash
+gemini extensions install https://github.com/Heretek-AI/IUMBTEMS --path plugins/gemini
+# local dev: gemini extensions link ./plugins/gemini
+```
+Bundle: `plugins/gemini/gemini-extension.json` (MCP servers, incl. the canonical `iumbtems` server), `GEMINI.md` (context), `commands/*.toml` (`/swarm`, `/grill`, `/audit`, `/scout`, `/brainstorming`, `/swarm-config`), `hooks/hooks.json`, `skills/` thin stubs.
+
+### 5. Codex CLI
+Codex reads repo-local `.agents/skills/*` (already generated in this repo) — `$brainstorming <prompt>` or implicit activation. Distributable pack in `plugins/codex/` (`openai.yaml`, `config.toml.snippet` for `[mcp_servers.*]`, `AGENTS.md.snippet`). Legacy `~/.codex/prompts/*.md` (`/prompts:<name>`) is deprecated; use skills.
+
+### 6. Google AntiGravity (`agy`)
+```bash
+agy plugin install https://github.com/Heretek-AI/IUMBTEMS --path plugins/antigravity
+agy plugin validate ./plugins/antigravity
+```
+Bundle: `plugins/antigravity/plugin.json`, `mcp_config.json`, `hooks.json`, `skills/` thin stubs, `agents/` (alpha/beta/auditor/brainstormer), `rules/epistemic-integrity.md`.
+
+### 7. OMP (oh-my-pi, `omp.sh`)
+```bash
+omp install npm:@heretek-ai/epistemic-swarm
+```
+Shares the pi extension entry point (`package.json: {omp, pi}` blocks). Project-local slash commands (`.omp/commands/*.md`), prompt templates (`.omp/prompts/*.md`), hooks (`.omp/hooks/pre|post/*.ts`), system override (`.omp/SYSTEM.md`).
+
 ---
 
 ## ⚡ CLI Command Reference
@@ -117,6 +143,9 @@ iumbtems scout "Zero-dependency Raft consensus implementations in Rust"
 # Run the autonomous dialectic research swarm
 iumbtems run "Sub-millisecond ZK state updates on L1 rollups"
 
+# Run lateral brainstorming (divergent what-if ideation, never bug fixes)
+iumbtems brainstorm "Where do we go from here?"
+
 # Run Socratic grilling and decision frontier calculation
 iumbtems grill --objective "L1 vs L2 state verification trade-offs"
 

package/bin/cli.js

@@ -2,7 +2,7 @@
 
 /**
  * IUMBTEMS CLI: I Use My Brain To Express My Self
- * High-Integrity Dialectic Research Harness for Claude Code, Pi, and OpenCode V2.
+ * High-Integrity Dialectic Research Harness for Claude Code, Pi, OMP, OpenCode V2, Gemini CLI, Codex CLI, AntiGravity.
  */
 
 const { spawnSync } = require('child_process');
@@ -17,7 +17,7 @@ function printHelp() {
   console.log(`
 🧠 IUMBTEMS: I Use My Brain To Express My Self
 🌟 Epistemic Swarm: High-Integrity Research Harness
-   Platforms: Claude Code | Pi (pi.dev) | OpenCode V2
+   Platforms: Claude Code | Pi (pi.dev) | OMP (oh-my-pi) | OpenCode V2 | Gemini CLI | Codex CLI | AntiGravity
    Organization: @heretek-ai | Repository: Heretek-AI/IUMBTEMS
 
 Usage:
@@ -29,7 +29,9 @@ Commands:
   audit "<target>"      Run dialectic codebase architectural & security audit
   scout "<feature>"     Scout open-source software, libraries & clean-room blueprints
   run "<objective>"     Run the dialectic multi-agent research swarm
+  brainstorm "<prompt>"  Run lateral brainstorming (feature vectors + spikes)
   grill                 Launch interactive Socratic decision tree framing
+  adapters              Rebuild harness adapter mirrors (skills -> plugins/*, .agents)
   install               Install skills & MCP servers into ~/.claude/
   marketplace           Show Claude Code marketplace catalog & install commands
   test                  Run test suite for hashing, state machine & auditor
@@ -38,13 +40,17 @@ Commands:
 
 Platform Extensions:
   Pi (pi.dev):          pi install npm:@heretek-ai/epistemic-swarm
+  OMP (oh-my-pi):       omp install npm:@heretek-ai/epistemic-swarm
   OpenCode V2:          add "@heretek-ai/epistemic-swarm" to opencode.json plugins
+  Gemini CLI:           gemini extensions install https://github.com/Heretek-AI/IUMBTEMS --path plugins/gemini
+  Codex CLI:            copy plugins/codex/skills/* into .agents/skills/
+  AntiGravity:          agy plugin install https://github.com/Heretek-AI/IUMBTEMS --path plugins/antigravity
   Claude Code:          npx @heretek-ai/epistemic-swarm install
 
 Options:
   --mock-claude         Run swarm with synthetic mock responses (zero API cost)
   --frontier <file>     Path to settled frontier.json from Socratic grilling
-  --mode <mode>         Operating mode (research, audit, scout, hybrid)
+  --mode <mode>         Operating mode (research, audit, scout, hybrid, brainstorm)
   --engine <engine>     Search engine (duckduckgo, brave, firecrawl, searxng)
   --depth <n>           Max dialectic iterations (1-4)
   --dir <path>          Path to .research workspace directory (default: .research)
@@ -54,6 +60,7 @@ Examples:
   iumbtems audit "runner/ and skills/ concurrency & security"
   iumbtems scout "Zero-dependency Raft consensus in Rust"
   iumbtems run "Verify sub-millisecond ZK prover latency"
+  iumbtems brainstorm "Where do we go from here?"
   iumbtems grill --objective "Rollup architecture trade-offs"
   iumbtems doctor
 `);
@@ -122,11 +129,28 @@ switch (command) {
     break;
   }
 
+  case 'brainstorm': {
+    const objective = args[1];
+    if (!objective || objective.startsWith('--')) {
+      console.error('Error: Please provide a brainstorm prompt.');
+      console.error('Example: npx @heretek-ai/epistemic-swarm brainstorm "Where do we go from here?"');
+      process.exit(1);
+    }
+    const forwardArgs = ['--mode', 'brainstorm', '--objective', objective, ...args.slice(2)];
+    runPython('runner/research_swarm.py', forwardArgs);
+    break;
+  }
+
   case 'grill': {
     runPython('skills/grilling/socratic_tree.py', args.slice(1));
     break;
   }
 
+  case 'adapters': {
+    runPython('scripts/build_adapters.py', args.slice(1));
+    break;
+  }
+
   case 'install': {
     runBash('install.sh', args.slice(1));
     break;

package/config/domain_packs/biopharma.json

@@ -0,0 +1,23 @@
+{
+  "pack_id": "biopharma",
+  "description": "Biopharma regulatory research constitution: preprints heavily downweighted, retracted findings zero-tolerance, predatory/mirror domains banned. For FDA-facing dossiers and clinical evidence reviews.",
+  "tier_weights": {
+    "__default__": 1.0,
+    "PEER_REVIEWED": 1.0,
+    "TECHNICAL_DOCUMENTATION": 0.8,
+    "PRIMARY_PRESS": 0.7,
+    "PREPRINT": 0.3,
+    "SECONDARY_MEDIA": 0.2
+  },
+  "neg_bonus": 0.5,
+  "reject_penalty": 2.5,
+  "accept_threshold": 0.7,
+  "banned_domains": [
+    "predatory-journal.example",
+    "retracted-mirror.example",
+    "banned-preprint-mirror.example"
+  ],
+  "mandatory_tags": ["VERIFIED", "INFERRED", "HYPOTHESIS", "NEGATIVE_KNOWLEDGE"],
+  "retraction_policy": "zero_tolerance",
+  "export_template": "regulatory"
+}

package/config/domain_packs/legal.json

@@ -0,0 +1,19 @@
+{
+  "pack_id": "legal",
+  "description": "Legal research constitution: primary sources (statutes, filings, opinions) weighted above commentary; unverified secondary reporting heavily penalized.",
+  "tier_weights": {
+    "__default__": 1.0,
+    "PEER_REVIEWED": 0.8,
+    "TECHNICAL_DOCUMENTATION": 0.7,
+    "PRIMARY_PRESS": 1.0,
+    "PREPRINT": 0.2,
+    "SECONDARY_MEDIA": 0.3
+  },
+  "neg_bonus": 0.5,
+  "reject_penalty": 2.5,
+  "accept_threshold": 0.65,
+  "banned_domains": ["undisclosed-sponsored-content.example"],
+  "mandatory_tags": ["VERIFIED", "INFERRED", "HYPOTHESIS", "NEGATIVE_KNOWLEDGE"],
+  "retraction_policy": "standard",
+  "export_template": "regulatory"
+}

package/config/domain_packs/quant.json

@@ -0,0 +1,19 @@
+{
+  "pack_id": "quant",
+  "description": "Quantitative finance constitution: backtest claims must be reproducible; vendor marketing downweighted; live-trading failure postmortems valued as negative knowledge.",
+  "tier_weights": {
+    "__default__": 1.0,
+    "PEER_REVIEWED": 0.9,
+    "TECHNICAL_DOCUMENTATION": 1.0,
+    "PRIMARY_PRESS": 0.6,
+    "PREPRINT": 0.5,
+    "SECONDARY_MEDIA": 0.2
+  },
+  "neg_bonus": 0.75,
+  "reject_penalty": 3.0,
+  "accept_threshold": 0.7,
+  "banned_domains": ["vendor-whitepaper-mirror.example"],
+  "mandatory_tags": ["VERIFIED", "INFERRED", "HYPOTHESIS", "NEGATIVE_KNOWLEDGE"],
+  "retraction_policy": "standard",
+  "export_template": "regulatory"
+}

package/config/mcp-research-servers.json

@@ -1,5 +1,12 @@
 {
   "mcpServers": {
+    "iumbtems": {
+      "command": "python3",
+      "args": ["runner/mcp_server.py"],
+      "env": {
+        "PYTHONPATH": "${IUMBTEMS_ROOT}"
+      }
+    },
     "brave-search": {
       "command": "npx",
       "args": ["-y", "@modelcontextprotocol/server-brave-search"],