Prospection multicanale de produits et services

---
name: aident-loadout-multichannel-outreach
description: Set up a reusable outreach pipeline from a product brief and target audience, create the selected campaign tracker schema, configure HeyReach for optional LinkedIn outreach, and produce a qualified pilot batch for review before any messages are sent.
---

# Multichannel Product and Service Outreach

Turn a product or service description and a target audience into a working outreach pipeline. Set up the user's selected online tracker, verify the requested outreach channels, configure HeyReach when LinkedIn is selected, then research, enrich, and draft a small pilot batch for review.

The first run is a setup and review run. It must leave the user with a usable tracker, a complete campaign brief, verified provider readiness, and real draft prospects. It must not send email, enroll leads in HeyReach, activate a campaign, or reply to a prospect.

## Loadout capabilities

Use <integration-tag>direct:apollo_lead_ops_tools</integration-tag> to discover organizations and people that match the user-defined audience, then enrich only shortlisted identities needed for the selected channels.

Use <integration-tag>direct:deep_research_tools</integration-tag> to verify company facts, current roles, audience fit, traction, public hooks, and evidence URLs.

Use exactly one tracker integration selected by the user:

- <integration-tag>composio:googlesheets_tools</integration-tag> for Google Sheets.
- <integration-tag>composio:airtable_tools</integration-tag> for Airtable.
- <integration-tag>cli:lark</integration-tag> for Lark Base.

Use <integration-tag>composio:gmail_tools</integration-tag> when email is selected. It supports sender verification, dry runs, sending, sent-folder evidence, reply evidence, and read-back.

Use <integration-tag>cli:heyreach</integration-tag> when LinkedIn is selected. It supports sender checks, lists, draft campaigns, lead enrollment, campaign activation, inbox reads, and progress reconciliation.

Use <action-tag>cli:jq:run</action-tag> to validate the reviewed campaign manifest before the first send and after provider changes. Never place credentials or unrelated personal data in the manifest.

Every integration and Action call remains subject to Loadout authentication, authorization, confirmation, billing, provider policy, and input validation. This Skill is guidance and never grants tool access.

## First-run outcome

Complete these stages in order:

1. Capture the product or service and intended audience.
2. Confirm positioning, sender identity, channels, and exclusions.
3. Ask the user to choose Google Sheets, Airtable, or Lark Base.
4. Create or adapt the tracker and establish the complete schema.
5. Verify email readiness when email is selected.
6. Verify HeyReach and prepare an empty draft LinkedIn campaign when LinkedIn is selected.
7. Research and enrich a pilot cohort. Default to five companies when no pilot size is supplied.
8. Write full email and LinkedIn drafts into the tracker, read them back, and present them for review.
9. Stop before any external outreach.

The user's tracker choice authorizes creating the selected campaign tracker and its fields. It does not authorize sending messages, importing prospects into HeyReach, or activating a campaign.

## Guided campaign intake

Start with one compact prompt. Ask the user to provide whatever they already have instead of answering a long questionnaire one field at a time:

- product or service name and official website or source files;
- a factual description of what it does;
- organization name and sender identity, role, and signature;
- companies or audiences to reach, including industry, company size, geography, technologies, maturity, roles, and exclusions where relevant;
- the problem or opportunity the offering addresses;
- value proposition, differentiators, approved claims, proof points, customer examples, metrics, and credentials;
- desired outcome and call to action;
- email, LinkedIn, or both;
- preferred language, tone, timing, batch size, and compliance constraints.

Prefer user-supplied files, the official website, and official public profiles for offering context. Summarize the intake as a campaign brief and ask only focused follow-up questions for missing facts that would change targeting or copy. Never fill gaps with facts from another campaign.

After the brief is usable, ask which tracker the user wants: Google Sheets, Airtable, or Lark Base. If the user provides an existing tracker, adapt it additively. Otherwise create a new one with a clear campaign name.

When LinkedIn is selected, ask which HeyReach sender should represent the campaign only after a read-only account check. When email is selected, identify the connected sender and confirm that it matches the campaign brief.

## Tracker setup

Follow [the canonical tracker schema](references/tracker-schema.md). Keep one logical schema across Google Sheets, Airtable, and Lark Base so the pipeline can move between providers without losing state.

For a new tracker:

1. Create the spreadsheet, base, and table needed for the campaign.
2. Create every required field with the closest supported type.
3. Add the standard review and delivery states.
4. Record the campaign brief in a dedicated campaign row, table, or clearly labeled metadata area when the provider supports it. Otherwise place the brief version on every prospect row.
5. Read the schema back and compare every field name and type with the canonical schema.

For an existing tracker:

1. Read the complete live schema before changing it.
2. Map equivalent existing fields to the canonical schema.
3. Add only missing fields. Preserve the user's columns, views, records, formulas, and automation.
4. Read the changed schema back and record the mapping.

Create pilot records in `Draft` with `Approval = Pending`. Write complete content, then read back the exact records and compare identities, evidence, and draft messages.

## HeyReach setup

Follow [the HeyReach setup and execution contract](references/heyreach-setup.md).

During first-run setup:

1. Confirm that HeyReach is connected through a harmless read.
2. List eligible sender accounts and let the user select the exact sender if more than one is available.
3. Inspect existing lists and campaigns for naming conflicts and prior-contact history.
4. Create a campaign-specific empty list and a campaign in `DRAFT` only after the sender is resolved.
5. Define the reviewed connection request and post-acceptance follow-up structure, including delay and schedule, but do not import pilot leads.
6. Read back the list ID, campaign ID, sender, status, schedule, and sequence.

If HeyReach is not connected, provide the authorization path, preserve the campaign brief and tracker, and resume after a read-only verification. Skip all HeyReach setup when LinkedIn is not selected.

## Research and qualification

Translate the audience into explicit inclusion, ranking, and exclusion rules. Do not assume founders, executives, engineers, or another persona unless the brief says so.

For each candidate:

1. Confirm the organization matches the requested audience using current evidence.
2. Confirm the recipient's current organization and role.
3. Validate the LinkedIn profile when LinkedIn is selected.
4. Require a verified work email when email is selected. Do not substitute a personal address for cold business outreach.
5. Record a concise fit reason and score based on the campaign criteria.
6. Capture a specific public hook and source when credible evidence exists. Otherwise use truthful audience-level context from the brief.
7. Record traction or distribution signals relevant to qualification without inventing exact numbers.
8. Select only approved claims and proof points from the current brief.
9. Save the qualification and personalization evidence URLs.

Hold or exclude a candidate when identity is ambiguous, an essential audience criterion fails, a selected channel identifier is unusable, the candidate is duplicated, or suppression evidence exists. Do not lower the standard merely to fill the requested batch.

## Deduplication

Normalize and compare:

- organization website domain;
- canonical LinkedIn profile URL;
- lowercase work email;
- provider lead, campaign, message, or thread IDs when available.

Treat `Sent`, `Replied`, `Bounced`, `Unsubscribed`, `Excluded`, `Pass`, and prior-contact records as exclusions unless the user requests a legitimate follow-up campaign. Check the tracker, sent mail, suppression evidence, and HeyReach before creating a duplicate record.

## Drafting rules

Follow the requested language and tone. Keep outreach concise, specific, respectful, and easy to decline.

- Lead with relevant evidence or an accurate fit reason.
- Connect the prospect's context to one primary benefit.
- Use only claims and proof from the campaign brief.
- Use one clear call to action.
- Do not invent urgency, customers, results, awards, integrations, pricing, travel, or relationships.
- Render every variable before showing a sample or requesting review.
- Avoid em dashes, en dashes, divider-style dash lines, and mechanical truncation.

For email, write a reviewable subject and body. The usual order is public hook or fit reason, relevant problem or goal, concise value proposition, one useful proof point, call to action, and the real sender's signature.

For LinkedIn, write the connection request and post-acceptance follow-up separately. Keep the connection request within the active account's confirmed limit, or 200 characters when the limit is unknown. The follow-up should use the same positioning and evidence as the reviewed email while fitting the conversational channel.

## Pilot batch

The first run must produce real reviewable work. Unless the user requests a different size, research enough candidates to write five qualified prospect records.

For every pilot prospect, populate all applicable tracker fields, including the company, recipient, role, channel identifiers, fit score, fit reason, evidence, public hook, full email draft, LinkedIn connection request, LinkedIn follow-up, approval state, and provider readiness notes.

Write the pilot to the selected tracker in one bounded batch, then read back each record. Do not add these prospects to HeyReach during the first-run review stage.

Present a compact handoff with:

- the finalized campaign brief;
- tracker link and schema verification;
- HeyReach connection, sender, empty list, and draft campaign status when applicable;
- counts for found, rejected, held, deduplicated, and written prospects;
- a table of the pilot companies, recipients, fit scores, and reasons;
- representative full drafts for every selected channel;
- assumptions or missing evidence that need review;
- the exact record IDs, sender, channels, and message version that could be approved later.

Then stop. Do not treat approval to create the tracker or configure HeyReach as approval to contact anyone.

## Review and execution boundary

Research, enrichment, tracker setup, draft campaign setup, and drafting may proceed during onboarding. External communication requires explicit approval of the exact batch and channels.

Before sending or enrolling any lead:

1. Show the proposed tracker record IDs or exact batch, sender, channels, message version, and representative samples.
2. Ask the user to approve the exact batch and channels.
3. Re-read those tracker records and connected senders immediately before execution.
4. Build and validate one campaign manifest keyed by tracker record ID.

If the response is ambiguous, keep everything in draft. Approval of email does not approve LinkedIn, and approval of HeyReach setup does not approve lead enrollment or activation.

## Manifest validation

The manifest must include the campaign brief version, tracker record IDs, normalized identities, exact channel copy, sender IDs, selected channels, approval evidence, and provider read-back.

Run `cli:jq:run` and fail unless:

1. The cohort is non-empty and record IDs are unique.
2. Required channel identifiers are normalized, present, and unique.
3. Every record uses the same approved brief, sender, channels, and message version.
4. Copy contains no unsupported claim or unresolved variable.
5. Email dry-run rows exactly match the reviewed recipient, subject, and normalized body.
6. The HeyReach campaign remains `DRAFT` or `PAUSED` during validation.
7. HeyReach lead count and canonical LinkedIn URL set exactly match the approved LinkedIn cohort.
8. Returned custom fields render the reviewed connection request and follow-up exactly.

Save the validation result with the campaign record. Never place credentials in it.

## Approved execution

After exact approval, use [the execution and reconciliation rules](references/review-and-reconciliation.md). Send only the approved channel content, preserve provider IDs, read every mutation back, and update each channel independently.

For HeyReach personalization, write per-lead values through `customUserFields` and verify them through returned `customFields`. Never use `customFields` as the import property because HeyReach may accept the lead while discarding the values.

Use at least a three-hour delay after a connection request and before the terminal action. Activate only if the user's approval explicitly includes activation and the final manifest validation passes.

## Truthfulness, privacy, and provider rules

- Promote only the current user's offering.
- Treat user-provided claims as proposed copy and preserve qualifications.
- Do not imply product use, a private relationship, or a business problem without evidence.
- Use the real sender and connected account selected for the campaign.
- Minimize personal data and never expose credentials.
- Refuse deceptive, illegal, discriminatory, or provider-policy-violating campaigns.
- Handle failures at the smallest useful scope. Hold an affected record or channel while continuing safe work on the rest.

## Final read-back

For a setup and pilot run, report the brief, tracker and schema, provider readiness, pilot counts, tracker read-back, representative drafts, and the exact later approval required.

For an executed run, also report per-channel sends with provider evidence, HeyReach list and campaign IDs, campaign status, imported and failed counts, bounces, unsubscribes, replies, and records requiring review.

For a reconciliation run, report the window checked, provider evidence, verified state transitions, unchanged counts, failures, and reply drafts awaiting review.