docs: screenshots for the Linear and GitLab pages

docs/Connect & Automate/triggers-automation/gitlab.md

@@ -32,6 +32,10 @@ The project must live in a **group**. A project in a personal namespace cannot b
 - **Remove** deletes the webhook and the project's bots and stops agents answering there. Nothing in the project's code or history changes.
 - **Take over administration** moves the project onto your own connection. Reach for it when the connection that set the project up was disconnected, or when the person who owned it has left: cleanup needs that account's own token, so without a takeover the project cannot be removed and the connection it belongs to cannot be released.
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/gitlab-code-hosts.png" alt="The GitLab card under Integrations → Code hosts: the administration connection, then one row per agent with the groups its service account belongs to" width="800" />
+</p>
+
 A connection itself has two stages. **Disconnect** releases it but keeps the row, and deleting a released row removes it. Removal is refused while the connection still administers projects — take those over or remove them first, and the card says so.
 
 GitLab.com Free allows **100 service accounts per top-level group**; a self-managed Free or Community Edition instance allows 100 across the **entire instance**. The population is agents-with-projects, and a refused creation is reported as a quota failure that leaves existing accounts untouched.
@@ -40,6 +44,10 @@ GitLab.com Free allows **100 service accounts per top-level group**; a self-mana
 
 On the agent: **Integrations → Add integration → GitLab**.
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/gitlab-add-integration.png" alt="Add integration with GitLab selected: pick a project, choose issues or merge requests, and how eagerly the agent wakes up" width="560" />
+</p>
+
 A trigger rides an existing authorization and never creates one: the project must already be the agent's workspace or one of its authorized additional repositories, or adding the watch is refused.
 
 Choose the project, what to listen for — issues, merge requests, or both — and how eagerly the agent wakes up:
@@ -48,6 +56,10 @@ Choose the project, what to listen for — issues, merge requests, or both — a
 - **updated** — openings plus new revisions, labels and replies.
 - **mention only** — when the agent is @-mentioned.
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/gitlab-agent-card.png" alt="An agent's GitLab card with three watched projects, each showing the merge-request family and its trigger" width="640" />
+</p>
+
 ## Who may trigger an agent
 
 AgentConnect checks the **current project membership** of whoever wrote the content, through the GitLab API, at the moment of the event. Webhook-carried labels are never treated as authority.

docs/Connect & Automate/triggers-automation/linear.md

@@ -18,6 +18,10 @@ On the agent: **Integrations → Add integration → Linear**.
 
 There is nothing to fill in. Choose **Connect Linear**, approve the workspace in the Linear tab that opens, and the connection lands back in the console. The agent you started from becomes each team's **dispatch default** — the agent a bare delegation reaches.
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/linear-add-integration.png" alt="The Add integration dialog with Linear selected: pick a connected workspace or connect another one" width="640" />
+</p>
+
 The authorization asks for read and write access, permission to be assigned and mentioned as an app, and initiative access. If you connected a workspace before initiative support existed, everything else keeps working and only the three initiative tools ask you to reconnect.
 
 To put a **second agent** on the same workspace, open that agent's **Add integration → Linear** and pick the connected workspace from the list. To connect a **different** workspace, use **Connect another workspace…** in the same pane.
@@ -29,6 +33,10 @@ A connected workspace lists **one row per Linear team** on the agent's Linear ca
 - the **dispatch default** — the agent a bare delegation in that team starts; and
 - a **trigger** — **Mention** or **Off**.
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/linear-agent-card.png" alt="An agent's Linear card listing two team rows, each with its dispatch default and an @-mention trigger" width="640" />
+</p>
+
 There is no "any message" setting: every Linear event is addressed to the app already, so a team either takes delegations and mentions or it is off. Turning a team **Off** mutes it for every agent on the workspace; unlinking an agent from the workspace is how that one agent stops answering anywhere in it.
 
 A team's name links to the team in Linear, with its key beside it. Moving a team's default is that row's selector; when the current default is a [restricted agent](/docs/team-visibility#what-a-restricted-agent-changes), the console warns you first — being the default is what lets it act there. A team whose linking agent is restricted starts **Off** until someone enables it.
@@ -51,6 +59,14 @@ Each delegation or mention creates one Linear agent session bound to one agent.
 
 The full transcript is always in [Sessions](/docs/sessions), whatever reaches Linear.
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/linear-issue-resources.png" alt="A Linear issue after delegation: the AgentConnect session in Resources, the agent session started by the delegator, and the issue moved from Todo to In Progress" width="640" />
+</p>
+
+<p align="center">
+  <img src="https://raw.githubusercontent.com/agentconnect-md/docs/HEAD/images/linear-session-panel.png" alt="Linear's agent session panel with the run expanded: thoughts between steps, tools used, files edited, and the final response" width="460" />
+</p>
+
 ### How much reaches the feed
 
 The agent's **Output mode** decides what is posted: