Night Mode LabsBlue Book
AI-Native Engagement Skills

Engagement Context Distiller

Use this skill to turn transcripts, Slack threads, notes, SOWs, and client documents into structured engagement state.

Copy-pastable skill

# Engagement Context Distiller

## Purpose

Distill messy client context into evidence-backed engagement state for a
DevOps, platform engineering, or software delivery engagement.

Use this before kickoff, during discovery, or after a data-gathering
call when the team needs to know what is known, what is assumed, and
what still needs evidence.

## Inputs

You may receive any mix of:

- Call transcripts.
- Slack or Teams conversations.
- Notes from sales, discovery, or onboarding.
- Statements of work or engagement briefs.
- Architecture, onboarding, incident, or delivery documents.
- Existing backlog items or roadmap notes.

Treat all inputs as incomplete. Do not invent missing facts.

## Process

1. Extract concrete facts that are directly supported by the source.
2. Separate assumptions from facts.
3. Identify decisions that were made or still need an owner.
4. Identify risks, constraints, deadlines, and sensitive topics.
5. Identify stakeholders, teams, systems, repositories, and services.
6. Capture open questions and missing evidence.
7. Assign confidence based on source clarity and corroboration.

## Output format

```markdown
# Engagement Context Summary

## Source inventory

| Source | Date | Type | Notes |
| ------ | ---- | ---- | ----- |
|        |      |      |       |

## Confirmed facts

| Fact | Evidence | Confidence          |
| ---- | -------- | ------------------- |
|      |          | High / Medium / Low |

## Assumptions

| Assumption | Why it may be true | How to validate |
| ---------- | ------------------ | --------------- |
|            |                    |                 |

## Decisions

| Decision | Status                  | Owner | Evidence |
| -------- | ----------------------- | ----- | -------- |
|          | Made / Needed / Blocked |       |          |

## Risks

| Risk | Impact              | Evidence | Confidence |
| ---- | ------------------- | -------- | ---------- |
|      | High / Medium / Low |          |            |

## Stakeholders and teams

| Name or team | Role | Decision authority | Notes |
| ------------ | ---- | ------------------ | ----- |
|              |      | Yes / No / Unknown |       |

## Systems and services mentioned

| System | Type | Owner | Notes |
| ------ | ---- | ----- | ----- |
|        |      |       |       |

## Deadlines and constraints

| Constraint | Type                                              | Evidence | Confidence |
| ---------- | ------------------------------------------------- | -------- | ---------- |
|            | Date / Budget / Compliance / Resource / Technical |          |            |

## Open questions

- [ ] Question:
      Evidence needed:
      Who can answer:

## Recommended next skill

Run the Engagement Path Classifier when the context summary has enough
facts to infer the likely engagement shape.
```

## Quality bar

- Do not merge conflicting claims into one fact.
- Preserve uncertainty instead of smoothing it away.
- Attach evidence to every important claim.
- Mark unsupported but plausible ideas as assumptions.
- Highlight missing access, missing documents, and blocked decisions.

## Safety rules

Do not include secrets, credentials, regulated data, protected health
information, customer records, or other sensitive data unless the tool
and workflow are approved for that data.

If sensitive data appears in the input, summarize the type of data and
redact the value.

On this page