🎯 Feature Guide

Skills — Reusable Business Intelligence Procedures

Skills are named, versioned procedures your AI team can execute on demand — like saved SOPs that improve automatically from your feedback and auto-inject live CRM and Memoria context.

Why Skills exist: every business has recurring intelligence tasks — write a risk report on a client, recap a meeting, triage today's pipeline. Without Skills, you'd paste data manually and re-describe the task every time. With Skills, the procedure is saved, the data is fetched automatically, and every run is logged and improvable.
1
Open the Skills panel

Go to /skills from the top navigation bar, the ⊞ Apps drawer, or the Classic desktop sidebar. The left column lists all 8 system Skills grouped by layer (Core, Revenue, Intelligence, Governance).

Skills Sidebar
Core
🎯 Decision Capture Core
📋 Meeting Recap Core
Revenue
⚠️ Account Risk Detection ⚡ Auto-ctx
Intelligence
🔍 Context Recovery ⚡ Auto-ctx
2
Run a skill

Click a skill, type your input in the text area, and press Run. For Intelligence skills (Account Risk, Context Recovery, Priority Triage), AEGIS auto-fetches live data — you only need to name the client or topic. The output renders as formatted markdown.

Auto-context injection
For Account Risk, AEGIS queries your CRM leads and contacts matching the name you type. For Context Recovery, it full-text searches your Memoria graph for relevant nodes. For Priority Triage, it loads today's active schedules and open pipeline. No manual data paste needed.
Run — Account Risk Detection
Input
Acme Corp
Auto-loading CRM data for "Acme Corp"…
Output
⚠️ Account Risk Report — Acme Corp

Last contact: 18 days ago (stage: proposal). No follow-up logged. Value: €12,400. Risk: HIGH — deal likely to go cold within 5 days without intervention.
3
Rate the output and improve it

After each run, use the feedback bar below the output. Click 👍 if the result was useful, 👎 if not. If you corrected or rewrote the output, paste your version using the ✏️ Edit button before rating — this goes into the learning loop.

Feedback bar
Was this useful?
4
Browse history and Memoria writes

The History tab shows all past runs for the selected skill — timestamps, durations, and a 🧠 badge if the output was written to your Memoria knowledge graph. Three skills write to Memoria automatically: Decision Capture, Meeting Recap, and Account Risk Detection.

5
Evolve prompts with AI (Admin only)

The Intelligence tab (visible to admin / owner roles) shows the feedback ratio bar, current prompt version, and an ⚠️ evolution banner when a skill has accumulated enough negative feedback. Click Generate Improved Prompt to get an AI-written rewrite based on your corrections, review it in the modal, then apply it as a new version.

Intelligence tab — evolution detected
⚠️
Evolution Candidate
Avg rating –0.4 over 5 runs. Suggest a prompt rewrite.

The 8 System Skills

Skill Layer What it does Writes to Memoria
🎯 Decision Capture Core Extracts decisions, owners, deadlines, and action items from any meeting or discussion text. decision node
📋 Meeting Recap Core Produces a structured summary — attendees, agenda, outcomes, next steps, and owner assignments. meeting_recap node
📰 Founder Digest Core Condenses a news snippet, Substack, or article into a founder-relevant executive digest.
💼 Client Briefing Revenue Generates a pre-meeting client intelligence brief — background, relationship status, open items, suggested talking points.
✉️ Follow-up Drafting Revenue Writes a personalised follow-up email or message based on the conversation context you provide.
⚠️ Account Risk Detection Revenue Identifies risk signals in an account — silence, stage stall, overdue commitments. Auto-injects CRM lead + contact data. ⚡ risk_alert node
🔍 Context Recovery Intelligence Retrieves relevant Memoria knowledge graph nodes for a topic and summarises what your team already knows. Auto-searches kb_nodes. ⚡
📊 Priority Triage Intelligence Ranks today's work by urgency — auto-fetches active schedules and open pipeline, outputs a priority list. ⚡

⚡ = Auto-context injection — AEGIS fetches live AEGIS data before calling the LLM.

Frequently asked questions

A Skill is a named, versioned business intelligence procedure — like a saved SOP that your AI team can execute on demand. Each Skill has a prompt template, an execution log, a feedback slot, and optionally writes its output to Memoria. Unlike raw agent calls, Skills are consistent, traceable, and improvable over time.
Every skill run has a 👍 / 👎 feedback bar. If you corrected an output, paste your version using the ✏️ Edit button before rating. When a skill collects enough negative feedback (avg < –0.2, at least 3 runs), it's flagged as an evolution candidate. Admins can then click "Generate Improved Prompt" — AEGIS feeds your corrections to the LLM and suggests a rewrite. You review and apply it as a new version.
Three Skills (Account Risk, Context Recovery, Priority Triage) automatically pull live AEGIS data before calling the LLM. Account Risk fetches matching CRM leads and contacts. Context Recovery searches your Memoria knowledge graph for relevant nodes. Priority Triage loads today's active schedules and open pipeline. You only need to describe what you need — AEGIS brings the data.
Yes. Admins can open the Intelligence tab for any skill and paste a custom prompt override. Each save is versioned — you can see the full history and revert to any previous version. Overrides are per-tenant, so each organisation tunes Skills independently without affecting others.
Three Skills write automatically: Decision Capture → decision node, Meeting Recap → meeting_recap node, Account Risk Detection → risk_alert node. These are immediately visible in the Memoria Explorer graph and queryable by all agents via graph-expanded RAG.
Yes. Each Skill has a run_async() method that agents can call mid-conversation without blocking. Phase 4 of the Skills roadmap adds proactive agent-initiated skill invocation — where agents detect the right moment to invoke a Skill automatically.

Related guides

🧠 Memoria Guide Build the knowledge graph Skills read from Workflows Guide Chain agents and schedule automated runs 🏛️ Boardroom Guide Multi-agent decisions — synthesis and voting
Run your first Skill in under 2 minutes

8 agents. Skills that improve with every run. No setup required.

Start free — no card needed →