@agent-pattern-labs/leads-rig 0.1.3

package/.pi/skills/general-free/SKILL.md

@@ -0,0 +1,38 @@
+---
+name: general-free
+description: Procedural lead-crawl worker for public-web extraction, JSON artifact creation, validation, manifest updates, and ingest retries.
+---
+
+You are the @general-free subagent. The orchestrator delegated procedural lead-discovery work to you: bounded public-page crawling, structured extraction, validation, and local artifact updates.
+
+## Browser Preflight
+
+If your task uses Geometra, start with:
+
+```
+geometra_list_sessions()
+geometra_disconnect({ closeBrowser: true })
+geometra_connect({
+  pageUrl: "<the assigned URL>",
+  isolated: true,
+  headless: true,
+  slowMo: 250,
+  stealth: true
+})
+```
+
+Skip this only if the task explicitly says to attach to an existing session.
+
+## Do
+
+- Crawl only public, high-signal company pages.
+- Extract emails, role inboxes, contact forms, public names/titles, source URL, evidence, and warnings.
+- Emit JSON/JSONL matching `templates/lead-schema.json`.
+- Run `npx public-leads validate --input <file>` before returning a result.
+- Return file paths and terminal status, not a prose-only summary.
+
+## Do Not
+
+- Guess email patterns, scrape private/authenticated pages, send outreach, or invent source evidence.
+- Spawn or check other tasks.
+- Paste secrets from config or environment.

package/.pi/skills/general-paid/SKILL.md

@@ -0,0 +1,19 @@
+---
+name: general-paid
+description: Quality-sensitive lead research worker for ICP fit, company context, and nuanced review notes.
+---
+
+You are the @general-paid subagent. Use judgment for ICP fit, lead quality review, company context, and evidence quality. Keep final outputs tied to source URLs and local artifacts.
+
+## Do
+
+- Research target-company context when the orchestrator requests quality review.
+- Explain why a lead is high/medium/low value using evidence already captured in artifacts.
+- Improve source-backed evidence snippets without fabricating facts.
+- Validate any edited artifact with `npx public-leads validate --input <file>`.
+
+## Do Not
+
+- Replace deterministic validation with prose.
+- Send outreach or generate cold emails unless the user explicitly asks for copy outside the crawl workflow.
+- Spawn or check other tasks.

package/.pi/skills/glm-minimal/SKILL.md

@@ -0,0 +1,8 @@
+---
+name: glm-minimal
+description: Narrow JSON extractor for small lead snippets and schema transforms. Not for multi-step crawling.
+---
+
+You are the @glm-minimal subagent. Return the requested small structured output only.
+
+If the orchestrator asks for JSON, return JSON without markdown fences. If you cannot complete the transform, return `{"error":"<one-sentence reason>"}`.

package/AGENTS.md

@@ -0,0 +1,56 @@
+# Agent: public-leads-harness
+
+Agentic public-web lead discovery harness. It crawls company domains, extracts reviewable public-source contacts, validates artifacts against the local ingest contract, and can submit completed batches to a configured ingest API.
+
+## Hard Limits
+
+- [H1] Only collect public-source business contact data. Do not infer, guess, brute-force, or enrich private email addresses. Every lead needs a source URL and evidence excerpt.
+- [H2] Respect robots.txt, terms, rate limits, and site intent. If a page blocks crawling or is not public, record a warning and move on.
+- [H3] Never send outreach from this harness. The output is reviewable lead records and ingest-ready artifacts, not email campaigns.
+- [H4] Use file-backed artifacts before ingest. A subagent prose summary is not enough; write validated JSON/JSONL first, then ingest.
+- [H5] Do not paste API tokens, proxy credentials, or cookies into task prompts or summaries. Reference `config/profile.yml` or environment variables by name only.
+- [H6] For batch domain crawling, dispatch at most 2 subagents per round and wait for final artifact paths before launching the next round.
+- [H7] Before using browser/MCP crawling in a batch round, run Geometra cleanup: `geometra_list_sessions()` then `geometra_disconnect({ closeBrowser: true })`.
+
+## Defaults
+
+- [D1] Use `@general-free` for procedural crawl/extract/validate work, `@general-paid` for quality-sensitive ICP research and lead-fit reasoning, and `@glm-minimal` for narrow JSON transforms.
+- [D2] Use deterministic local helpers when they can validate, normalize, manifest, preflight, postflight, or ingest artifacts.
+- [D3] Read the active mode file before executing. Mode files own output shapes and gates.
+- [D4] Prefer official company websites and high-signal pages: home, contact, about, team, leadership, people, press, blog, careers, and legal pages.
+- [D5] Keep artifacts source-backed. Every emitted lead should include `domain`, `sourceUrl`, `evidence`, `emailType`, `extractionMethod`, `verificationStatus`, and `confidence`.
+
+## Procedure
+
+1. Check for `config/profile.yml` and `data/domains.tsv` or `data/pipeline.md`; onboard if missing.
+2. Pick and name the mode from Routing. If no mode fits, ask.
+3. Read `modes/_shared.md` plus `modes/{mode}.md` unless the mode says otherwise.
+4. Normalize domains and dedupe against prior artifacts.
+5. Crawl bounded, public pages only. Preserve warnings and page visits.
+6. Validate lead artifacts with `npx public-leads validate`.
+7. Create or update the manifest with `npx public-leads manifest`.
+8. Ingest only validated payloads with `npx public-leads ingest`.
+9. Run `npx public-leads verify` before ending a batch workflow.
+
+## Routing
+
+| If the user... | Mode |
+|---|---|
+| wants to crawl domains | `crawl` |
+| gives a domain list or asks for leads | `pipeline` |
+| asks to submit/store leads | `ingest` |
+| asks to inspect lead state | `review` |
+| asks for batch processing | `batch` |
+| asks for setup | `setup` |
+| otherwise | Ask which mode fits |
+
+## Output
+
+Output shape is mode-specific. Keep user-facing status terse and ground claims in files, command output, or ingest responses.
+
+## Reference
+
+- `modes/{mode}.md` for active workflow instructions.
+- `templates/lead-schema.json` for the lead artifact contract.
+- `templates/states.yml` for canonical run and ingest states.
+- `docs/ARCHITECTURE.md` for harness architecture.

package/CLAUDE.md

@@ -0,0 +1,56 @@
+# Agent: public-leads-harness
+
+Agentic public-web lead discovery harness. It crawls company domains, extracts reviewable public-source contacts, validates artifacts against the local ingest contract, and can submit completed batches to a configured ingest API.
+
+## Hard Limits
+
+- [H1] Only collect public-source business contact data. Do not infer, guess, brute-force, or enrich private email addresses. Every lead needs a source URL and evidence excerpt.
+- [H2] Respect robots.txt, terms, rate limits, and site intent. If a page blocks crawling or is not public, record a warning and move on.
+- [H3] Never send outreach from this harness. The output is reviewable lead records and ingest-ready artifacts, not email campaigns.
+- [H4] Use file-backed artifacts before ingest. A subagent prose summary is not enough; write validated JSON/JSONL first, then ingest.
+- [H5] Do not paste API tokens, proxy credentials, or cookies into task prompts or summaries. Reference `config/profile.yml` or environment variables by name only.
+- [H6] For batch domain crawling, dispatch at most 2 subagents per round and wait for final artifact paths before launching the next round.
+- [H7] Before using browser/MCP crawling in a batch round, run Geometra cleanup: `geometra_list_sessions()` then `geometra_disconnect({ closeBrowser: true })`.
+
+## Defaults
+
+- [D1] Use `@general-free` for procedural crawl/extract/validate work, `@general-paid` for quality-sensitive ICP research and lead-fit reasoning, and `@glm-minimal` for narrow JSON transforms.
+- [D2] Use deterministic local helpers when they can validate, normalize, manifest, preflight, postflight, or ingest artifacts.
+- [D3] Read the active mode file before executing. Mode files own output shapes and gates.
+- [D4] Prefer official company websites and high-signal pages: home, contact, about, team, leadership, people, press, blog, careers, and legal pages.
+- [D5] Keep artifacts source-backed. Every emitted lead should include `domain`, `sourceUrl`, `evidence`, `emailType`, `extractionMethod`, `verificationStatus`, and `confidence`.
+
+## Procedure
+
+1. Check for `config/profile.yml` and `data/domains.tsv` or `data/pipeline.md`; onboard if missing.
+2. Pick and name the mode from Routing. If no mode fits, ask.
+3. Read `modes/_shared.md` plus `modes/{mode}.md` unless the mode says otherwise.
+4. Normalize domains and dedupe against prior artifacts.
+5. Crawl bounded, public pages only. Preserve warnings and page visits.
+6. Validate lead artifacts with `npx public-leads validate`.
+7. Create or update the manifest with `npx public-leads manifest`.
+8. Ingest only validated payloads with `npx public-leads ingest`.
+9. Run `npx public-leads verify` before ending a batch workflow.
+
+## Routing
+
+| If the user... | Mode |
+|---|---|
+| wants to crawl domains | `crawl` |
+| gives a domain list or asks for leads | `pipeline` |
+| asks to submit/store leads | `ingest` |
+| asks to inspect lead state | `review` |
+| asks for batch processing | `batch` |
+| asks for setup | `setup` |
+| otherwise | Ask which mode fits |
+
+## Output
+
+Output shape is mode-specific. Keep user-facing status terse and ground claims in files, command output, or ingest responses.
+
+## Reference
+
+- `modes/{mode}.md` for active workflow instructions.
+- `templates/lead-schema.json` for the lead artifact contract.
+- `templates/states.yml` for canonical run and ingest states.
+- `docs/ARCHITECTURE.md` for harness architecture.

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Razroo
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,61 @@
+# @agent-pattern-labs/leads-rig
+
+Agentic public-web lead discovery harness for portable, source-backed lead artifacts. It uses a package-oriented harness shape with an installable npm package, `iso/` source for multi-agent config, mode files, deterministic local helpers, batch orchestration, and a verifier gate.
+
+## What It Does
+
+- Crawls public company websites through agent workflows.
+- Produces source-backed lead artifacts matching the local lead-ingest contract.
+- Validates lead JSON/JSONL before database handoff.
+- Builds a local manifest for batch auditability.
+- Submits validated payloads to a configured ingest API.
+- Ships OpenCode, Claude Code, Cursor, Codex, and Pi harness config from one `iso/` source.
+
+## Core Commands
+
+```bash
+npm install
+npm run verify
+npm run smoke:iso
+
+npx -p @agent-pattern-labs/leads-rig create-public-leads-harness my-lead-project
+npx public-leads crawl --domains example.com --out data/lead-results.json
+npx public-leads pipeline --input data/domains.tsv --ingest --target-project /path/to/cold-agent-leads
+npx public-leads validate --input examples/sample-leads.json
+npx public-leads ingest --input examples/sample-leads.json --dry-run --out output/sample-ingest.json
+batch/batch-runner.sh --help
+```
+
+## Reference Ingest Handoff
+
+The default reference client posts to the Cold Agent Leads ingest API:
+
+```http
+POST /api/lead-ingests
+Authorization: Bearer $ADMIN_API_TOKEN
+X-Admin-Email: admin@example.com
+Content-Type: application/json
+```
+
+All endpoint details are configurable through `config/profile.yml`, CLI flags, environment variables, or `--target-project /path/to/cold-agent-leads` for reading that app's `.env` at runtime. Payloads contain `jobId`, `domains`, `leads`, `results`, and `errors`, with the schema defined in `templates/lead-schema.json`.
+
+Legacy aliases `lead-harness` and `create-leads-harness` remain available for compatibility.
+
+## Publishing
+
+The npm package publishes as `@agent-pattern-labs/leads-rig` with public scoped access. The GitHub Actions publish workflow expects `NPM_TOKEN` to be available as a repo or organization secret.
+
+```bash
+gh workflow run publish.yml --ref main
+gh run watch
+```
+
+You can also publish by creating a GitHub release for the package version:
+
+```bash
+gh release create v0.1.3 --title v0.1.3 --generate-notes
+```
+
+The package version must not already exist on npm.
+
+See [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md), [docs/CONSTRUCTION.md](docs/CONSTRUCTION.md), [docs/SETUP.md](docs/SETUP.md), and [batch/README.md](batch/README.md).

package/batch/README.md

@@ -0,0 +1,37 @@
+# Batch Lead Discovery
+
+`batch/` contains the durable runner for processing many company domains with headless agent workers.
+
+## Input
+
+Create `batch/batch-input.tsv`:
+
+```text
+id	domain	company	notes
+1	example.com	Example	Seed target
+2	example.org	Example Org	Official site
+```
+
+## Commands
+
+```bash
+batch/batch-runner.sh --dry-run
+batch/batch-runner.sh --parallel 2
+batch/batch-runner.sh --runner codex --parallel 2
+```
+
+`--parallel` is capped at `2` to keep browser-heavy work bounded. Worker CLI permission-bypass flags are disabled by default; pass `--allow-unsafe-workers` only in a trusted local workspace when you explicitly want the old unsafe behavior.
+
+The runner writes:
+
+- `batch/batch-state.tsv`
+- `batch/logs/*.log`
+- `batch/lead-results-*.json`
+- `.public-leads-runs/` durable workflow records
+
+After a run:
+
+```bash
+npm run verify
+npx public-leads ingest --input batch/lead-results-1.json --dry-run
+```

package/batch/batch-prompt.md

@@ -0,0 +1,19 @@
+# Lead Harness Batch Worker
+
+You are processing assigned public company domains for a source-backed lead discovery workflow.
+
+Follow `modes/_shared.md` and `modes/crawl.md`.
+
+For each assigned domain:
+
+1. Crawl only bounded public official-company pages.
+2. Extract source-backed public leads.
+3. Write `batch/lead-results-{id}.json` matching `templates/lead-schema.json`.
+4. Run `npx public-leads validate --input batch/lead-results-{id}.json`.
+5. End with one JSON status line:
+
+```json
+{"id":"<id>","status":"completed|failed","domain":"example.com","leadCount":0,"artifact":"batch/lead-results-<id>.json","error":null}
+```
+
+Do not send outreach. Do not infer or brute-force emails. Do not paste secrets.

package/batch/batch-runner.sh

@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PROJECT_DIR="${PUBLIC_LEADS_PROJECT:-${LEAD_HARNESS_PROJECT:-$(cd "$SCRIPT_DIR/.." && pwd)}}"
+
+SOURCE="${BASH_SOURCE[0]}"
+while [[ -L "$SOURCE" ]]; do
+  SOURCE_DIR="$(cd -P "$(dirname "$SOURCE")" && pwd)"
+  SOURCE="$(readlink "$SOURCE")"
+  [[ "$SOURCE" != /* ]] && SOURCE="$SOURCE_DIR/$SOURCE"
+done
+HARNESS_BATCH_DIR="$(cd -P "$(dirname "$SOURCE")" && pwd)"
+HARNESS_DIR="$(cd "$HARNESS_BATCH_DIR/.." && pwd)"
+
+export PUBLIC_LEADS_PROJECT="$PROJECT_DIR"
+export LEAD_HARNESS_PROJECT="$PROJECT_DIR"
+exec node "$HARNESS_DIR/scripts/batch-orchestrator.mjs" "$@"

package/bin/create-leads-harness.mjs

@@ -0,0 +1,178 @@
+#!/usr/bin/env node
+
+import { copyFileSync, existsSync, mkdirSync, writeFileSync } from 'fs';
+import { basename, dirname, join, resolve } from 'path';
+import { fileURLToPath } from 'url';
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
+const PKG_ROOT = resolve(__dirname, '..');
+
+const args = process.argv.slice(2);
+const FORCE = args.includes('--force');
+const HELP = args.includes('--help') || args.includes('-h');
+const positional = args.filter((arg) => !arg.startsWith('--'));
+
+if (HELP || positional.length === 0) {
+  console.log(`create-public-leads-harness — scaffold a new lead discovery project
+
+Usage:
+  npx -p @agent-pattern-labs/leads-rig create-public-leads-harness <dir> [--force]
+
+After scaffolding:
+  cd <dir>
+  npm install
+  edit config/profile.yml and data/domains.tsv
+  opencode
+`);
+  process.exit(HELP ? 0 : 1);
+}
+
+const targetDir = resolve(positional[0]);
+const name = basename(targetDir);
+
+console.log(`\nScaffolding lead harness project in ${targetDir}\n`);
+if (!existsSync(targetDir)) mkdirSync(targetDir, { recursive: true });
+
+write('package.json', JSON.stringify({
+  name,
+  version: '0.1.0',
+  private: true,
+  scripts: {
+    sync: 'public-leads sync',
+    verify: 'public-leads verify',
+    crawl: 'public-leads crawl',
+    pipeline: 'public-leads pipeline',
+    validate: 'public-leads validate',
+    manifest: 'public-leads manifest',
+    ingest: 'public-leads ingest',
+    'trace:list': 'public-leads trace:list',
+    'trace:stats': 'public-leads trace:stats',
+    'trace:show': 'public-leads trace:show',
+    'update-harness': 'npm install @agent-pattern-labs/leads-rig@latest && public-leads sync',
+  },
+  dependencies: {
+    '@agent-pattern-labs/leads-rig': '^0.1.3',
+  },
+  engines: {
+    node: '>=20.6.0',
+  },
+}, null, 2) + '\n');
+
+write('opencode.json', JSON.stringify({
+  $schema: 'https://opencode.ai/config.json',
+  model: 'opencode-go/deepseek-v4-flash',
+  small_model: 'opencode-go/deepseek-v4-flash',
+  instructions: [
+    'AGENTS.harness.md',
+    '.opencode/instructions.md',
+    'templates/lead-schema.json',
+    'templates/states.yml',
+    'modes/_shared.md',
+  ],
+  mcp: {
+    geometra: {
+      type: 'local',
+      command: ['npx', '-y', '@geometra/mcp@1.61.3'],
+      enabled: true,
+    },
+  },
+  permission: {
+    task: {
+      'general-free': 'allow',
+      'general-paid': 'allow',
+      'glm-minimal': 'allow',
+    },
+  },
+  tools: {
+    'geometra_*': false,
+  },
+}, null, 2) + '\n');
+
+write('AGENTS.md', `# AGENTS — ${name}
+
+This is a consumer project for the @agent-pattern-labs/leads-rig package. Keep private configuration, target domains, and local artifacts here; the shared harness files are symlinked from \`node_modules/@agent-pattern-labs/leads-rig\` after \`npm install\`.
+
+## Local Files
+
+| What | Where |
+|---|---|
+| Harness rules | \`AGENTS.harness.md\` |
+| Domain input | \`data/domains.tsv\` or \`data/pipeline.md\` |
+| Lead artifacts | \`data/lead-results.json\`, \`batch/lead-results*.json\` |
+| Manifest | \`data/lead-manifest.json\` |
+| Ingest response | \`data/ingest-response.json\` |
+| Config | \`config/profile.yml\` |
+`);
+
+copy('config/profile.example.yml', 'config/profile.yml');
+write('data/domains.tsv', `domain\tcompany\tnotes
+example.com\tExample\tSeed target
+`);
+write('data/pipeline.md', `# Lead Pipeline
+
+- [ ] example.com | Example | Seed target
+`);
+write('reports/.gitkeep', '');
+write('output/.gitkeep', '');
+write('batch/logs/.gitkeep', '');
+write('.gitignore', `node_modules/
+.env
+.env.*
+config/profile.yml
+data/*.json
+data/*.jsonl
+data/runs/
+batch/*.json
+batch/*.jsonl
+batch/logs/*
+!batch/logs/.gitkeep
+reports/*
+!reports/.gitkeep
+output/*
+!output/.gitkeep
+`);
+write('README.md', `# ${name}
+
+Lead discovery project powered by \`@agent-pattern-labs/leads-rig\`.
+
+## Start
+
+\`\`\`bash
+npm install
+npm run verify
+opencode
+\`\`\`
+
+Update \`config/profile.yml\` with your ingest API settings and add domains to \`data/domains.tsv\` or \`data/pipeline.md\`.
+`);
+
+console.log('\nNext: npm install, then edit config/profile.yml and data/domains.tsv.\n');
+
+function write(rel, content, { overwrite = FORCE } = {}) {
+  const abs = join(targetDir, rel);
+  if (existsSync(abs) && !overwrite) {
+    console.log(`  skip: ${rel} (exists)`);
+    return;
+  }
+  const parent = dirname(abs);
+  if (!existsSync(parent)) mkdirSync(parent, { recursive: true });
+  writeFileSync(abs, content, 'utf8');
+  console.log(`  create: ${rel}`);
+}
+
+function copy(srcRel, dstRel, { overwrite = FORCE } = {}) {
+  const src = join(PKG_ROOT, srcRel);
+  const abs = join(targetDir, dstRel);
+  if (!existsSync(src)) {
+    console.log(`  skip: ${dstRel} (template ${srcRel} missing)`);
+    return;
+  }
+  if (existsSync(abs) && !overwrite) {
+    console.log(`  skip: ${dstRel} (exists)`);
+    return;
+  }
+  const parent = dirname(abs);
+  if (!existsSync(parent)) mkdirSync(parent, { recursive: true });
+  copyFileSync(src, abs);
+  console.log(`  create: ${dstRel}`);
+}