Salesforce Query Openclaw

Salesforce Query (Schema-Adaptive)

Most Salesforce scripts work in one org, then break in the next. This skill was built to fix that.

It discovers your schema first, then runs reusable GTM workflows against what actually exists in your environment.

Source code

• GitHub: https://github.com/RikH81/salesforce-query-openclaw-skill

Security model

• Credentials are stored in macOS Keychain (openclaw.salesforce)
• No plaintext credential-file fallback
• Supports session-only mode via --no-save
• Restricts SALESFORCE_INSTANCE_URL to HTTPS Salesforce domains

1) Configure credentials

Required variables:

• SALESFORCE_CLIENT_ID
• SALESFORCE_CLIENT_SECRET
• SALESFORCE_INSTANCE_URL

Preferred flow:

1. Create a Salesforce Connected App and dedicated least-privilege integration user.
2. Store secrets in a secure secret manager if possible.
3. Default persistence is macOS Keychain. Plaintext credential-file fallback has been removed.

Credentials can be persisted locally via:

from salesforce_client import save_credentials, check_credentials

ok, missing = check_credentials()
if not ok:
    save_credentials(client_id="...", client_secret="...", instance_url="...")

2) Run onboarding discovery

Run an end-to-end discovery pass:

python3 scripts/onboarding.py

For session-only credentials (nothing persisted):

python3 scripts/onboarding.py --no-save

If credentials are missing, onboarding prompts for them and saves them before continuing.

If credentials are not ready, run quick setup guidance mode:

python3 scripts/onboarding.py --guided

You can also type help at the first credential prompt to show setup guidance inline.

If Python dependency requests is missing, onboarding will stop and print explicit setup commands for a local .venv-salesforce.

Run a credential safety check any time:

python3 scripts/credential_doctor.py

For automation or CI, pass credentials directly:

python3 scripts/onboarding.py \
  --client-id "$SALESFORCE_CLIENT_ID" \
  --client-secret "$SALESFORCE_CLIENT_SECRET" \
  --instance-url "$SALESFORCE_INSTANCE_URL" \
  --non-interactive

Session-only credentials (no disk persistence):

python3 scripts/onboarding.py \
  --client-id "$SALESFORCE_CLIENT_ID" \
  --client-secret "$SALESFORCE_CLIENT_SECRET" \
  --instance-url "$SALESFORCE_INSTANCE_URL" \
  --non-interactive --no-save

Optional custom output path:

python3 scripts/onboarding.py --out ~/.config/openclaw/salesforce_profile.json

This will:

1. Validate auth.
2. Discover accessible objects.
3. Describe fields on core GTM objects.
4. Detect optional signal packs (intent, revenue, risk, engagement).
5. Generate targeted follow-up questions.
6. Save a reusable profile JSON.

3) Ask adaptive follow-ups

Use generated questions from onboarding output, then persist answers in the profile.

Reference question bank:

• references/question-bank.md

4) Build and use mapping profile

Use assets/profile-template.json as the canonical profile format.

Keep uncertain mappings marked as low confidence until user confirms.

5) Query patterns

For ready-to-use workflows, use:

python3 scripts/starter_workflows.py

or import functions from scripts/starter_workflows.py:

• top_accounts_to_work_today(...)
• pipeline_rescue_board(...)
• campaign_followup_opportunities(...)

Each returns a structured payload with summary, rows, and next_best_action to proactively guide users.

Use salesforce_client.py for regular querying:

from salesforce_client import SalesforceClient

sf = SalesforceClient()
accounts = sf.get_account_by_name("Acme")
opps = sf.get_open_opportunities()

For schema discovery:

from salesforce_client import SalesforceClient
from schema_discovery import SchemaDiscovery

sf = SalesforceClient()
discovery = SchemaDiscovery(sf)
schema = discovery.discover()

References

• Discovery approach: references/discovery-playbook.md
• Follow-up prompts: references/question-bank.md
• Security and credential setup: references/security-setup.md
• Starter workflow guide: references/starter-workflows.md