Claude Code builds artifact-grounded sourcing briefs
Lemieux reports a Claude Code recruiting workspace that turns public candidate artifacts into dossiers and specific outreach anchors, with reusable tone and data rules.
Original prompt
Shared by the source author. See source notes for context.
Create a short candidate dossier from public artifacts and extract 2-3 outreach angles.
Reach people exploring AI agents
See placements →What happened
Lemieux reports a Claude Code recruiting workspace that turns public candidate artifacts into dossiers and specific outreach anchors, with reusable tone and data rules.
Reported by Sophie LemieuxHow the agent did it
- Set role personas, tone rules and data boundaries in the workspace.
- Supply a public candidate artifact to the dossier skill.
- Extract concrete work, trade-offs and relevant outreach angles.
Where a person stepped in
Lemieux defines persona and communication rules and tracks response quality. Candidate personal information is excluded from the repository; autonomous sending is not documented.
Recruiter Sophie Lemieux describes a Claude Code workspace used in her daily sourcing work. A persistent instruction file stores her recruiting principles, tone and measurement conventions. Separate files hold persona templates, outreach examples and benchmark definitions, while the repository excludes candidate personal information.
Her dossier skill accepts a public artifact such as a GitHub profile, repository, conference talk or long post. It produces a compact summary of what the person built, their trade-offs, stack and scope, then extracts outreach angles and a specific reason for contact. A related skill turns one researched observation into a first message: an observation, a connection to the role and a low-friction question. Boolean-building and follow-up skills support the same workspace.
The concrete reported result is a reusable sourcing process with artifact-based dossiers and messages, rather than a measured improvement in recruiting conversions. The opening example's weak reply rate is background context, not proof that this agent fixed that funnel. Lemieux tracks reply quality and opportunity rate instead of trusting privacy-distorted email opens. The copyable detail is to require a defensible public observation before drafting contact and keep sensitive candidate records out of the reusable instruction repository. The source does not document autonomous sending or an independently validated candidate-ranking system.
Prompt
Original prompt
Source: Blog · Sophie Lemieux
Frequently asked questions
What grounds a first message?
A specific public artifact and a defensible observation about the person’s work, connected to the role.
Where is candidate personal information stored?
The described repository rules prohibit storing candidate personal information there; reusable process context is kept separate.
Does the account prove higher reply rates?
No post-change conversion benchmark is supplied. The author describes her working setup and the metrics she chooses to track.
More details
Remaining steps
- Draft a first message grounded in one observed artifact.
- Review the draft and measure reply quality rather than email opens.
About this account
Claude Code researches supplied artifacts, creates structured dossiers and drafts targeted messages using reusable workspace files and skills.
Firsthand account by Sophie Lemieux; original publication date verified against Substack byline Feb 20, 2026. Outcomes are the author’s reports, not independently reproduced. Original dossier skill description excerpt from the primary article; not the full skill instructions.
Credit: Sophie Lemieux
Source published 20 Feb 2026
Source checked 14 Oct 2026
Tools and topics
Claude Code · Public GitHub artifacts · Recruiting skills · Markdown templates
Candidate dossiers · Public-artifact sourcing · Data boundaries
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.