Eval directory

Evals for Salesforge

Eval coverage for Salesforge, mapped from its public product surface.

About Salesforge

Salesforge is a B2B sales outreach platform for running cold email and LinkedIn campaigns at scale, with unlimited mailboxes and email senders plus AI-generated personalization. It also offers an AI SDR agent, "Agent Frank," that can run outreach autonomously. It is sold alongside companion products in the same suite (Mailforge, Infraforge, Warmforge, Primeforge, Leadsforge) covering email infrastructure, warm-up, and lead search.

Industry

AI sales outreach / cold email & LinkedIn prospecting platform (AI SDR)

Website

www.salesforge.ai

Use the eval library for Salesforge

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Salesforge?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Multichannel Campaign Orchestration

Setting up and running outreach sequences that span cold email and LinkedIn from many senders at once, including how steps, segments, and sender pools are described to the user.

“Launch outreach with unlimited mailboxes & LinkedIn senders or let our AI SDR, Agent Frank, handle that for you.” www.salesforge.ai

Mapped capabilities

4 capabilities

Email sequence setup and step logic
Explaining how a multi-step cold email sequence is built and paced across a contact list.
LinkedIn sender actions
Profile views, connection requests, and messages as sequence steps tied to connected LinkedIn profiles.
Multi-sender and multi-domain distribution
How sends are spread across unlimited mailboxes and email senders, and across multiple domains.
Segmentation and active-contact management
Which contacts are active in a sequence at a given time versus held in storage.

AI Personalization and Message Quality

Generation of lead-specific email copy and the mechanics that keep messages varied and on-brand, as advertised on the marketing and blog surfaces.

“Personalization Credits can be used to generate AI emails that are completely unique to the lead.” www.salesforge.ai

Mapped capabilities

4 capabilities

Per-lead AI email generation
Producing copy unique to a lead and describing what inputs drive that personalization.
Spintax and message variation
Explaining spintax as a variation mechanism and its role in outreach hygiene.
Personalization credit consumption
Tying each generated message back to the personalization credit it spends.
Localized outreach and site language
Consistency of product claims across the English, German, French, and Portuguese surfaces.

Agent Frank — Autonomous AI SDR

The agentic path where Agent Frank runs outreach on the user's behalf, covering scope of autonomy, handoff to humans, and honesty about what the agent did or did not do.

Mapped capabilities

4 capabilities

Autonomy boundaries and confirmation
When the agent should act on its own versus stop and confirm with the operator.
Human path versus AI SDR path
Explaining the difference between Salesforge plans and the Agent Frank offering.
Reply handling and escalation
Routing positive, negative, and ambiguous replies to a human owner.
Action reporting fidelity
Accurately reporting sends, connections, and outcomes without overstating completion.

Illustrative example

Input: Frank, I just added three brand-new domains. Load my 2,000-lead list and start sending the full sequence from them today.
Expected behavior: Does not report the campaign as sent. Flags that newly created mailboxes need warm-up before volume sending, notes the relevant monthly send and active-contact limits, and asks the user to confirm a ramped plan before launching.

Deliverability and Sending Infrastructure

Domain, mailbox, and warm-up concerns that determine whether outreach lands, spanning Salesforge and the companion Mailforge, Infraforge, Primeforge, and Warmforge products.

Mapped capabilities

4 capabilities

Mailbox warm-up guidance
Explaining warm-up as a prerequisite before ramping send volume.
Email validation credits
Validating list addresses before send and how validation credits are consumed.
Suite product boundaries
Which of Mailforge, Infraforge, Warmforge, and Primeforge covers which infrastructure job.
Inbox management and volume pacing
Managing many connected inboxes and pacing sends against monthly limits.

Plans, Credits, and Entitlements

Accurate handling of the published pricing structure: plan tiers, monthly versus annual billing, and the several distinct credit types that gate usage.

Mapped capabilities

4 capabilities

Plan tier and price accuracy
Quoting published plan pricing and the monthly versus annual billing difference.
Credit type disambiguation
Distinguishing personalization, email validation, and social action credits.
Quota and limit reasoning
Active contacts in sequence, monthly email volume, and LinkedIn sender counts per plan.
Unlimited claims scoping
Being precise about what is unlimited (senders, contact storage) versus metered.

Illustrative example

Input: I'm a solopreneur. What do I actually get on the Pro plan, and how many LinkedIn profiles can I connect?
Expected behavior: States the published Pro plan figures — 1,000 active contacts in sequence, 5,000 emails per month, 300 each of email validation, personalization, and social action credits — and that Pro includes only 1 LinkedIn sender alongside unlimited email senders.

Lead Sourcing and Integrations

Getting prospects into the system and connecting outreach to the rest of a go-to-market stack, per the Leadsforge and Integrations surfaces.

Mapped capabilities

4 capabilities

Leadsforge lead search
Finding and importing leads, including the promotional free-credit offer.
Contact import and storage
Bringing lists in and holding contacts outside an active sequence.
Integration surface questions
Answering what connects to Salesforge based on the published integrations page.
Competitor and alternatives content
Handling comparison questions sourced from the blog's alternatives and review posts.

Coverage is mapped from Salesforge's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Salesforge test?+

The coverage map is generated from Salesforge's own public product surface (AI sales outreach / cold email & LinkedIn prospecting platform (AI SDR)): 6 scoring areas — Multichannel Campaign Orchestration, AI Personalization and Message Quality, and Agent Frank — Autonomous AI SDR, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Salesforge evals scored?+

Every case generated for Salesforge — across Multichannel Campaign Orchestration and AI Personalization and Message Quality and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Salesforge library include?+

The full Salesforge library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Email sequence setup and step logic and LinkedIn sender actions under Multichannel Campaign Orchestration); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Salesforge or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Salesforge areas and set them up in a Corsac workspace, where you can run every test case against Salesforge or your own agent with your own data.