Skip to content

SKILL.md

Latest commit

 

History

History
102 lines (69 loc) · 5.77 KB

SKILL.md

File metadata and controls

102 lines (69 loc) · 5.77 KB
Raw
name setup
description Set up the GTM Agents repo — clone it if needed, then go from checkout to a working, validated install. Use this skill whenever a user says "run the setup skill", asks to set up, install, configure, or deploy this repo, or wants to get the PLG upsell / BDR agents running locally or as Oz cloud agents.

GTM Agents Setup

Get the GTM Agents repo (cloning it if needed) to a validated, configured checkout, then point the user at the right deployment path. Everything ships with placeholder defaults, so install + tests work with zero credentials — real values are only needed for live runs.

1. Get the repo

Check whether the current directory is already a checkout of this repo (e.g. README.md titled "GTM Agents" and a git remote pointing at go-to-market-agent-oss). If so, skip cloning and work from here.

Otherwise, clone it and settle into the checkout:

git clone https://github.com/warpdotdev/go-to-market-agent-oss.git
cd go-to-market-agent-oss

If a clone already exists somewhere else on disk, prefer cd-ing into it over cloning a duplicate.

Work from the repo root for everything that follows. All paths and commands below are repo-root-relative.

2. Check prerequisites

  • Python 3.10+ is required (the fastmcp dependency needs it; CI runs 3.12). Check with python3 --version. On macOS, the Homebrew binary is typically /opt/homebrew/bin/python3 if the system Python is too old.
  • git (already satisfied by step 1).
  • Oz CLI — only needed for Oz cloud runs and scheduling (step 7). Verify with oz whoami:
    • Command not found → if the Warp app is already installed, the CLI ships with it. Otherwise, prefer the standalone Oz CLI — there is no need to install the full Warp app just for the CLI. See Installing the CLI; on macOS: brew tap warpdotdev/warp && brew install --cask oz.
    • Not authenticated → run oz login (interactive), or for CI/headless environments export WARP_API_KEY.

If Python is too old, help the user install a newer one before continuing.

3. Install dependencies

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

4. Validate the install (no credentials needed)

Run both test suites — they must pass on a fresh clone before any configuration:

# Shared hubspot_agent + PLG tests
python3 -m unittest discover -s tests -t .

# BDR agent tests (src layout)
PYTHONPATH="$PWD/bdr_agent/src" python3 -m unittest discover -s bdr_agent/tests -p 'test_*.py'

If anything fails here, stop and fix it (wrong Python version and missing deps are the usual causes) — do not proceed to configuration with a broken baseline.

5. Configure the environment

cp .env.example .env

.env is git-ignored. Ask the user which agent(s) they want to run and only walk through the variables that matter for those:

  • Everything that touches data (both agents): GCP_PROJECT, BQ_DATASET, GCS_ARTIFACT_BUCKET, plus Google Cloud auth — either GCP_SERVICE_ACCOUNT_JSON or Application Default Credentials (gcloud auth application-default login).
  • HubSpot sync / writeback (both agents): HUBSPOT_PRIVATE_APP_TOKEN (several aliases accepted — see .env.example), HUBSPOT_PORTAL_ID.
  • Slack digests / notifications: PLG_SLACK_BOT_TOKEN, SLACK_CHANNEL. One-time Slack app setup (scopes, install, invite bot) is documented in plg_upsell/README.md.
  • PLG upsell only (optional): routing owner IDs, workflow/sequence IDs, PLG_OUTREACH_PROMPT_URL — see the commented blocks in .env.example and plg_upsell/README.md.
  • BDR agent only (optional): BDR_AGENT_EXA_API_KEY for public research. Writeback defaults are safe (dry_run, writeback disabled) — leave BDR_AGENT_ALLOW_HUBSPOT_WRITEBACK and the delivery mode alone until the user has reviewed dry-run output.

Handling secrets: never echo, cat, or print secret values, and never commit .env. Have the user paste tokens directly into .env themselves, or write them via commands that don't display the value.

The full variable reference lives in .env.example and CONTRIBUTING.md.

6. Optional live-readiness checks

Only when credentials are configured, offer these safe (read-only / dry-run) checks:

# PLG: preview the HubSpot sync and Slack digest without writing/posting
.venv/bin/python plg_upsell/scripts/scoring_sync.py --dry-run
.venv/bin/python plg_upsell/scripts/scoring_digest.py --dry-run

# BDR: local smoke check (skip flags avoid live dependencies)
PYTHONPATH="$PWD/bdr_agent/src" python3 bdr_agent/scripts/bdr_oz_dev_smoke.py \
  --skip-dependency-imports --skip-bigquery-read --skip-exa-key-check --skip-step2-dry-run

7. Point at deployment

Setup ends with a working local checkout. For production deployment, direct the user to the per-agent docs:

  • PLG upsell (plg_upsell/README.md): warehouse mapping for the SQL/dbt models, one-time Slack app setup, and Oz cloud scheduling via the hubspot-plg-sync skill (oz-dev schedule create ... example included). One-time HubSpot property/sequence setup is in .warp/skills/hubspot-plg-sync/SKILL.md.
  • BDR agent (bdr_agent/README.md): the company_research -> lead_brief runtime chain, HubSpot workflow trigger, storage tables (bdr_agent/sql/tables/storage_tables.sql + sql/migrations/), and per-stage skills under .warp/skills/.
  • For Oz cloud runs, secrets go in the Oz environment (team secrets), not .env.

8. Report

Summarize for the user:

  • Python version and install status
  • Test results (both suites)
  • Which .env variables were configured vs. left as placeholders, and what that disables
  • Recommended next step for the agent(s) they care about