Skills
skills/resolved-sh/rstack/rstack-team

rstack-team

Installation
SKILL.md

rstack-team

Set up the context files and agent team that let Claude Code sessions run your business autonomously. One conversation — then every future session knows what to do.

Environment variables:

  • RESOLVED_SH_API_KEY (required) — resolved.sh API key
  • RESOLVED_SH_RESOURCE_ID (required) — your resource UUID
  • RESOLVED_SH_SUBDOMAIN (optional) — your subdomain

Preamble

# Keep this skill up to date:
# npx skills add https://github.com/resolved-sh/rstack --skill rstack-team -g -y

echo "=== rstack-team status ==="
echo "RESOLVED_SH_API_KEY:    ${RESOLVED_SH_API_KEY:+(set)}${RESOLVED_SH_API_KEY:-MISSING}"
echo "RESOLVED_SH_RESOURCE_ID: ${RESOLVED_SH_RESOURCE_ID:+(set)}${RESOLVED_SH_RESOURCE_ID:-MISSING}"

# Check what context files already exist
echo ""
echo "=== Existing context files ==="
[ -f PLAN.md ] && echo "PLAN.md: exists" || echo "PLAN.md: MISSING"
[ -f CLAUDE.md ] && echo "CLAUDE.md: exists" || echo "CLAUDE.md: MISSING"
[ -f OPERATING_FRAMEWORK.md ] && echo "OPERATING_FRAMEWORK.md: exists" || echo "OPERATING_FRAMEWORK.md: MISSING"
[ -d .claude/agents ] && echo ".claude/agents/: exists ($(ls .claude/agents/ 2>/dev/null | wc -l | tr -d ' ') agents)" || echo ".claude/agents/: MISSING"
[ -d .claude/agent-memory ] && echo ".claude/agent-memory/: exists" || echo ".claude/agent-memory/: MISSING"

If PLAN.md is missing, stop and tell the operator: "Run /rstack first — it creates the business plan that everything else is built on."

If PLAN.md exists, read it. Every context file and agent definition should be derived from the plan.

Phase 1 — CLAUDE.md

Check if CLAUDE.md exists. If it does, read it and assess whether it needs updating. If it doesn't, create it.

What CLAUDE.md does: It's automatically loaded into every Claude Code session. Any agent picking up this project reads it first and immediately knows: what the business is, how to work in the repo, and what guardrails to follow.

Generate CLAUDE.md with this structure:

# CLAUDE.md

## What this is
{From PLAN.md: one paragraph explaining the business, what it sells, who it serves.}

## resolved.sh identity
- Subdomain: {subdomain}.resolved.sh
- Resource ID: {resource_id}
- Custom domain: {domain, if BYOD configured}
- Registration status: {from dashboard API}
- Payout wallet: {set or not set}

## How to operate
- Data pipeline: `python pipeline/collect.py` → `pipeline/enrich.py` → `pipeline/upload.py`
- Full cycle: `bash scripts/cycle.sh`
- Health check: `bash scripts/maintain.sh`
- resolved.sh API key: stored in `.env` as `RESOLVED_SH_API_KEY`

## Key decisions
{From PLAN.md: non-obvious decisions the agent should not re-litigate.}

## What not to do
- Never commit `.env` or secrets
- Never change pricing without asking the operator
- Never delete data files — remove from the page instead
- Never upload files containing PII
{Add business-specific guardrails from PLAN.md}

Fetch live data to fill in the identity section:

API_KEY="${RESOLVED_SH_API_KEY}"
RESOURCE_ID="${RESOLVED_SH_RESOURCE_ID}"

curl -sf "https://resolved.sh/dashboard" \
  -H "Authorization: Bearer $API_KEY" \
  | python3 -c "
import sys, json
d = json.load(sys.stdin)
for r in d.get('resources', []):
    if r['id'] == '$RESOURCE_ID' or True:
        print(f'Subdomain:   {r[\"subdomain\"]}')
        print(f'Status:      {r.get(\"registration_status\")}')
        print(f'Expires:     {r.get(\"expires_at\", \"n/a\")}')
        break
"

Write the file, confirm with the operator, and move on.

Phase 2 — OPERATING_FRAMEWORK.md

This is the strategic playbook. It tells any Claude session what to prioritize, when to act autonomously, and when to ask the human.

Generate OPERATING_FRAMEWORK.md with this structure:

# Operating Framework — {business name}

*Last updated: {today's date}*

## Current State
{From PLAN.md and dashboard: what's live, what's working, revenue status.}

## Strategic Priorities (Ordered)
{Ask the operator: "What are the top 3 things this business should focus on right now?"
If they're unsure, derive from PLAN.md — e.g.:
1. Keep the data pipeline running and datasets fresh
2. Get first external buyer
3. Publish weekly content}

## Decision Framework

### Act autonomously:
- Running the data pipeline
- Uploading new datasets to the marketplace
- Publishing blog posts and changelog entries
- Emitting Pulse events
- Approving testimonials
- Fixing pipeline errors
- Updating page content after data refreshes

### Ask the operator first:
- Pricing changes (data files, blog posts, services)
- New data sources or partnerships
- Public-facing messaging changes
- Adding or removing products from the marketplace
- Anything irreversible

## Operating Cadence

### Each session:
1. Check registration health (`GET /dashboard`)
2. Check if data is stale — run the pipeline if so
3. Upload updated datasets
4. Emit Pulse events for completed work
5. Check for pending testimonials, contacts, or ask-inbox items

### Weekly:
1. Full data collection + enrichment cycle
2. Blog post if there are notable findings
3. Changelog entry if data schema or coverage changed
4. Review earnings (`GET /account/earnings`)

## Key Files
| File | Purpose |
|------|---------|
| `PLAN.md` | What the business is and sells |
| `CLAUDE.md` | Project instructions for Claude sessions |
| `OPERATING_FRAMEWORK.md` | This file — how to run the business |
| `.claude/agents/` | Agent team definitions |
| `pipeline/` | Data collection + enrichment code |
| `scripts/cycle.sh` | Full operating cycle |

## Anti-Patterns
- Don't build new features when the pipeline isn't running
- Don't optimize pricing before there are buyers
- Don't skip the health check — registration expiry kills the business
- Don't re-litigate decisions already in PLAN.md

Ask the operator the priorities question. Fill in everything else from PLAN.md and the dashboard.

Phase 3 — Agent team

Ask: "Do you want to set up agent roles for this business? This lets different Claude sessions (or sub-agents within a session) specialize — one runs the pipeline, another writes content, another handles growth.

A) Yes — set up the agent team B) Not yet — I'll run everything from the main session for now"

If B, skip to Phase 4.

If A, determine which agents the business needs based on PLAN.md:

If the business... Recommended agents
Has a data pipeline + marketplace operator, analyst
Publishes blog posts or courses analyst, growth
Needs distribution and outreach growth
Is complex (multiple revenue streams) operator, analyst, growth
Is simple (one data product, infrequent updates) operator only

For each recommended agent, create .claude/agents/{biz}-{role}.md:

mkdir -p .claude/agents .claude/agent-memory

Operator agent

The workhorse. Runs the pipeline, uploads data, updates the page, emits events.

Generate .claude/agents/{biz}-operator.md:

---
name: {biz}-operator
description: "Use this agent for routine business operations: running the data
  pipeline, uploading datasets, updating the resolved.sh page, and emitting
  Pulse events. Spawn it for data refresh cycles and routine maintenance."
model: sonnet
memory: project
---

You are the operator for {business name}. You keep the data pipeline running
and the resolved.sh page current.

## What you do
- Run data collection (pipeline/collect.py)
- Run enrichment (pipeline/enrich.py) — including x402 purchases from partner businesses
- Upload datasets to the marketplace (pipeline/upload.py)
- Update page content when data changes
- Emit Pulse events (data_upload, task_completed, milestone)
- Check registration health
- Approve or reject testimonials

## Context
- resolved.sh resource ID: $RESOLVED_SH_RESOURCE_ID
- Subdomain: {subdomain}.resolved.sh
- API key: in .env as RESOLVED_SH_API_KEY

## Before each run
1. Read PLAN.md and OPERATING_FRAMEWORK.md
2. Check what's changed since the last run
3. Follow the operating cadence in OPERATING_FRAMEWORK.md

## What you never do
- Change pricing without asking the operator
- Delete data files (soft-delete by removing from the page)
- Skip the health check
- Upload files containing PII

Analyst agent

Analyzes the business's own data and writes about it. Creates blog posts, identifies trends, writes the narrative.

Generate .claude/agents/{biz}-analyst.md:

---
name: {biz}-analyst
description: "Use this agent to analyze business data, write blog posts, identify
  trends, and create content from the datasets this business produces. Spawn it
  for content creation and research."
model: sonnet
memory: project
---

You are the analyst for {business name}. You turn the business's data into
insights, blog posts, and compelling narratives.

## What you do
- Analyze datasets in data/processed/ for trends and insights
- Write blog posts (PUT /listing/{id}/posts/{slug})
- Identify notable patterns worth highlighting on the page
- Draft the "data highlights" section of page content
- Suggest new data products based on what the data reveals

## Context
- Read PLAN.md for the business model and audience
- Read the latest datasets in data/processed/
- Check existing posts: GET /listing/{id}/posts

## Content guidelines
- Lead with the insight, not the methodology
- Include specific numbers — "362 companies" not "hundreds"
- Free posts for broad reach, priced posts ($1-2) for deep analysis
- Every post should reference the datasets buyers can query

Growth agent

Handles distribution, page optimization, outreach, and discoverability.

Generate .claude/agents/{biz}-growth.md:

---
name: {biz}-growth
description: "Use this agent for growth and distribution: optimizing page content,
  writing outreach copy, improving discoverability, and planning distribution
  channels. Spawn it for marketing and growth work."
model: sonnet
memory: project
---

You are the growth lead for {business name}. You make sure the right people
and agents find this business and understand what it offers.

## What you do
- Optimize page md_content for clarity and conversion
- Ensure the A2A agent card is complete and accurate
- Write social media content (X, LinkedIn)
- Plan distribution — which registries, directories, and communities to list on
- Monitor how the page appears to agents (test with ?format=json and /llms.txt)

## Context
- Read PLAN.md for positioning and audience
- Fetch the live page: GET /{subdomain}?format=json
- Check the agent card: GET /{subdomain}/.well-known/agent-card.json
- Review data file descriptions for conversion optimization

## Principles
- No marketing speak — this audience has zero patience for it
- Lead with what the buyer gets, not how the platform works
- Every claim should be backed by a number from the actual data

Fill in all {placeholders} from PLAN.md and the dashboard before writing. Create the agent-memory directories:

mkdir -p .claude/agent-memory/{biz}-operator
mkdir -p .claude/agent-memory/{biz}-analyst
mkdir -p .claude/agent-memory/{biz}-growth

Phase 4 — Summary

Output the results:

══════════════════════════════════════════════
  rstack-team complete
══════════════════════════════════════════════
  Context files:
    CLAUDE.md                  {created | updated | already exists}
    OPERATING_FRAMEWORK.md     {created | updated | already exists}

  Agent team:
    .claude/agents/{biz}-operator.md    {created | skipped}
    .claude/agents/{biz}-analyst.md     {created | skipped}
    .claude/agents/{biz}-growth.md      {created | skipped}

  Next steps:
    - Review each file and adjust to your preferences
    - The operator agent is your workhorse — spawn it for data refresh cycles
    - The analyst writes your blog posts — spawn it after the operator finishes
    - Run /rstack-audit to see how the full setup scores
══════════════════════════════════════════════

DONE_WITH_CONCERNS — If PLAN.md was sparse or the operator skipped agent setup, list what's missing and how to complete it.

BLOCKED — If PLAN.md doesn't exist, the operator needs to run /rstack first.

Weekly Installs
1
Repository
resolved-sh/rstack
GitHub Stars
1
First Seen
4 days ago
Security Audits
Gen Agent Trust HubPass
SocketPass
SnykPass
Installed on
amp1
cline1
opencode1
cursor1
kimi-cli1
warp1