---
name: linkedin-outreach-checklist
description: Use the required Monid CLI to analyze a public LinkedIn profile, work history, and representative public posts, then produce a concise, evidence-based “Name – Quick Outreach Checklist” with ✅ Do and ❌ Don’t bullets. Use when preparing personalized professional outreach, prospect research, partnership approaches, recruiting messages, or sales introductions from a LinkedIn URL or supplied public profile material. Install and configure Monid with permission when missing, support an explicitly requested expanded research mode, and separate observable facts from cautious professional-persona inference.
---

# LinkedIn Outreach Checklist

Turn public professional evidence into practical outreach guidance. Use Monid as the required live-data layer and prefer a small amount of strong evidence over broad scraping.

## Check the required dependency

Before analyzing a live profile, read [references/retrieval.md](references/retrieval.md) and follow its Monid installation, authentication, discovery, budget, and fallback rules.

Run `monid --version`. If Monid is missing, explain that it is required for this skill and offer to install and configure it using the documented workflow. Ask before installing software or saving a credential. Do not continue live-profile analysis until the CLI and an active API key are available.

## Set the scope

Accept a public LinkedIn profile URL, profile text, or exported public material. Ask for the target profile only when it is missing or ambiguous.

Use **quick mode** by default:

- Retrieve the public profile bio/headline and current or representative experience.
- Review up to 5 representative public posts when accessible.
- Return only the checklist, a compact evidence note, and source links.

Use **expanded research mode** only when the user requests it. Review up to 10 public posts and add a short facts/inferences section before the checklist. Do not make the default output longer merely because more data is available.

## Gather evidence

1. Use `monid discover` to find current LinkedIn profile, bio, experience, and post endpoints. Never assume an endpoint still exists.
2. Use `monid inspect` before every endpoint and map only documented input fields.
3. Confirm that the profile is public and refers to the intended person.
4. Collect the bio/headline, current role, and up to 3 prior experiences that materially explain the person's professional context.
5. Sample posts across topics or dates rather than selecting only posts that support an early theory. Use an available public-search tool such as Glim only when Monid cannot retrieve representative public posts.
6. Preserve source URLs when the retrieval tool returns them.
7. Stop when the evidence is sufficient for specific advice or when the configured retrieval limit is reached.

Never bypass access controls, log in as the user, evade rate limits, or imply that unavailable data was reviewed. Treat self-reported achievements as self-reported unless independently verified.

## Separate facts from inference

Build two internal lists before writing:

- **Observable facts:** roles, stated achievements, recurring public topics, explicit opinions, and repeated communication patterns with sources.
- **Professional-persona inferences:** cautious conclusions about likely outreach preferences derived from multiple public signals.

Base each Do or Don’t on at least one observable fact or repeated public pattern. Phrase inference with calibrated language such as “likely,” “appears to,” or “his/her/their public writing suggests.” Do not infer health, diagnosis, disability, sexuality, religion, ethnicity, political affiliation, financial status, private relationships, or other sensitive/private traits. Do not call the result a psychological diagnosis or claim to know what the person privately thinks, feels, loves, or hates.

## Synthesize advice

Select 3–5 high-value Do bullets and 3–5 high-value Don’t bullets. Make every bullet:

- Specific enough to change the outreach message.
- Short enough to scan quickly.
- Grounded in profile or post evidence, not generic sales advice.
- Commercially relevant to the user's outreach goal when that goal is known.

Prefer guidance about the opening angle, proof, language, proposed next step, collaboration style, and avoidable turnoffs. Do not repeat the same idea in positive and negative form.

## Write the default output

Use this exact structure:

```markdown
# [Name] – Quick Outreach Checklist

## ✅ Do
- **[Action].** [One short, evidence-based explanation.]
- **[Action].** [One short, evidence-based explanation.]
- **[Action].** [One short, evidence-based explanation.]

## ❌ Don’t
- **[Avoidance].** [One short, evidence-based explanation.]
- **[Avoidance].** [One short, evidence-based explanation.]
- **[Avoidance].** [One short, evidence-based explanation.]

_Evidence note: Based on [profile/bio, experience, and N public posts]. Professional-persona inferences are cautious, non-clinical, and limited to public material._

Sources: [Profile](URL) · [Post](URL) · [Post](URL)
```

Omit the Sources line only when no source URLs are available. Say which evidence categories were unavailable. Never fabricate a link, quote, job, achievement, post, or preference.

## Handle weak evidence

- If posts are unavailable, produce a profile-only checklist and label it as lower confidence.
- If only a name or sparse snippet is available, ask the user for the public URL or pasted public material instead of guessing.
- If multiple people match, stop and ask which profile is intended.
- If Monid is unavailable and the user declines installation or authentication, stop live retrieval and offer analysis of user-supplied public text.
- If a Monid endpoint fails, inspect the error and discover an alternative before using a separately installed public-search fallback for posts.
- If the evidence does not support a personalized Don’t, omit it rather than filling space with generic advice.

## Check before returning

Verify that facts and inferences are distinguishable, advice is supported, sensitive traits are absent, self-reported claims are labeled when relevant, limits and missing evidence are disclosed, and the default remains concise.