AI-Native Engagement Skills
Backlog Compiler
Use this skill to turn engagement artifacts, risks, and decisions into a ranked markdown backlog that can later be converted to Linear, Jira, or GitHub issues.
Copy-pastable skill
# Backlog Compiler
## Purpose
Compile a discrete, evidence-backed work backlog from engagement
context, discovery gaps, artifacts, risks, and decisions.
Use this when the team needs actionable work items without losing the
source evidence behind each task.
## Inputs
Use one or more of:
- Engagement context summary.
- Discovery gap report.
- Artifact pack.
- Risk register.
- Decision log.
- Current-state summary.
- Roadmap draft.
## Task types
Classify each item as one of:
- Discovery task.
- Delivery task.
- Decision task.
- Risk-reduction task.
- Handoff or adoption task.
- Validation or spike task.
## Process
1. Extract candidate work from risks, gaps, decisions, and roadmap
items.
2. Remove duplicates and merge overlapping tasks.
3. Split tasks that contain multiple outcomes.
4. Add acceptance criteria for each task.
5. Add source evidence and confidence.
6. Rank by risk reduction, urgency, dependency order, and value.
7. Mark tasks that need client decision or access before work starts.
## Output format
```markdown
# Initial Engagement Backlog
## Backlog summary
Primary outcome:
Ranking method:
Known blockers:
## Tasks
### Task 1: [Title]
**Type:** Discovery / Delivery / Decision / Risk reduction / Handoff /
Validation
**Why this matters:**
**Source evidence:**
- [Evidence reference]
**Acceptance criteria:**
- [ ] [Specific reviewable outcome]
- [ ] [Specific reviewable outcome]
- [ ] [Specific reviewable outcome]
**Dependencies:**
- [Dependency or none]
**Suggested owner role:**
**Priority:** High / Medium / Low
**Confidence:** High / Medium / Low
**Blocked by client action:** Yes / No
## Decisions needed before delivery
| Decision | Owner | Required by | Related tasks |
| -------- | ----- | ----------- | ------------- |
| | | | |
## Access or evidence blockers
| Blocker | Needed for | Owner | Priority |
| ------- | ---------- | ----- | -------- |
| | | | |
## Not creating tasks for
List any unsupported assumptions, vague wishes, or low-confidence ideas
that should not become backlog items yet.
```
## Quality bar
- Every task needs evidence or an explicit assumption.
- Acceptance criteria must be testable or reviewable.
- Do not create delivery tasks when discovery is still blocking scope.
- Prefer small tasks that can be completed in one focused session.
- Preserve dependency order.
- Flag tasks that require client action before work can start.