Skip to content
Back to skills

Making Decisions

ASecurity

Use when transforming synthesis insights into explicit decisions with documented trade-offs. Guides interactive decision-making and risk identification.

  • 7 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added May 27, 2026
data-aigodocumentation

Security analysis

A100/100

Pro scans all 3 files and shows the line behind each finding

Scanned May 27, 2026

npx -y skills add synaptiai/synapti-marketplace --skill making-decisions --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Making Decisions?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Making Decisions
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/synaptiai-making-decisions/badge)](https://www.skillsdirectory.com/skills/synaptiai-making-decisions)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: making-decisions
description: Use when transforming synthesis insights into explicit decisions with documented trade-offs. Guides interactive decision-making and risk identification.
context: fork
agent: general-purpose
---

# Decision Ledger

This skill guides the creation of explicit decisions with full trade-off documentation.

## Prerequisites

- Synthesis complete (`/ledger-synthesize`)
- `03-synthesis/CROSS-SYNTHESIS.md` exists with decision candidates
- Per-pillar syntheses in `03-synthesis/SYN-*.md`

## Workflow

Use TodoWrite to track these mandatory steps:

<required>
1. Load decision candidates from cross-synthesis
2. For each candidate, gather evidence and options
3. Present trade-offs for user decision
4. Create decision entry with semantic ID
5. Identify risks created by each decision
6. Generate DECISIONS.yaml and RISKS.yaml
7. Validate decision quality gates
</required>

### Step 1: Load Decision Candidates

Read `03-synthesis/CROSS-SYNTHESIS.md` to extract:
- Decision topics needing resolution
- Supporting evidence from each pillar
- Identified options

### Step 2: Gather Evidence per Decision

For each decision candidate:
- Collect all relevant EV-* IDs
- Summarize what evidence supports each option
- Note confidence levels

```yaml
decision_candidate:
  topic: target-segment-priority
  options:
    - name: SMB-first
      evidence:
        - EV-users-smb-pain-points (0.80)
        - EV-economics-smb-unit-economics (0.75)
    - name: Enterprise-first
      evidence:
        - EV-market-enterprise-tam (0.85)
        - EV-competitors-enterprise-gap (0.70)
```

### Step 3: Present Trade-offs

For each decision, present:
- Options with evidence support
- What you win with each option
- What you lose with each option
- Risks created by each option

Use **AskUserQuestion** to get user's decision.

### Step 4: Create Decision Entry

Write entry to `04-decisions/DECISIONS.yaml` using schema from [references/decision-ledger-schema.md](references/decision-ledger-schema.md).

<good-example>
```yaml
- id: DEC-scope-smb-first
  decision: "Target SMB segment before enterprise"
  status: accepted
  owner: user
  created_at: 2026-01-21
  alternatives:
    - Enterprise-first
    - Multi-segment simultaneous
  evidence:
    - EV-users-smb-pain-points
    - EV-economics-smb-unit-economics
    - EV-market-enterprise-tam
  tradeoffs:
    wins:
      - "Faster iteration cycles with smaller customers"
      - "Lower sales friction (self-serve possible)"
      - "Better unit economics at start"
    loses:
      - "Smaller initial contract values"
      - "May need significant pivot for enterprise later"
  risks:
    - RISK-market-smb-churn-rate
  implications:
    - "MVP UX must optimize for self-serve onboarding"
    - "Pricing must fit SMB budget constraints"
```
- Semantic ID (DEC-scope-smb-first)
- Explicit alternatives documented
- Multiple evidence citations (3+)
- Both wins AND loses documented
- Linked risks and implications
</good-example>

<bad-example>
```yaml
- id: DEC-001
  decision: "Do the thing"
  status: accepted
  alternatives: []
  evidence: []
  tradeoffs:
    wins:
      - "It will be good"
    loses: []
```
- Non-semantic ID (DEC-001)
- Vague decision statement
- No alternatives considered
- No evidence cited
- No loses documented (unrealistic)
</bad-example>

### Step 5: Identify Created Risks

Each decision may create risks. For each identified risk:
- Create entry in `05-risks/RISKS.yaml`
- Link back to creating decision
- Document triggers and mitigations

See [references/risk-ledger-schema.md](references/risk-ledger-schema.md) for schema.

### Step 6: Generate Ledger Files

Write complete files:
- `04-decisions/DECISIONS.yaml`
- `05-risks/RISKS.yaml`

### Step 7: Validate Quality Gates

**Decision quality gate:**
- Every decision cites ≥2 evidence IDs
- Every decision lists ≥1 alternative considered
- Every decision documents wins AND loses

**Risk quality gate:**
- Every risk links to creating decision
- Every risk has severity and likelihood
- Every risk has ≥1 mitigation

## User Interaction

Use the **AskUserQuestion tool** for every decision:

### Decision prompt
```
Question: "Decision needed: [topic]"
Options:
- "[Option A] - supported by [evidence summary]"
- "[Option B] - supported by [evidence summary]"
- "[Option C] - supported by [evidence summary]"
- "Need more information before deciding"
```

### Trade-off confirmation
```
Question: "You chose [option]. Confirming trade-offs:"
Options:
- "Yes, I accept these trade-offs"
- "Wait, I want to reconsider"
- "Explain the trade-offs more"
```

### Decision status
```
Question: "Should this decision be marked as:"
Options:
- "Accepted (committed)"
- "Provisional (may revisit)"
- "Need more research first"
```

### Risk severity
```
Question: "This decision creates risk: [risk]. How severe?"
Options:
- "High - requires immediate mitigation"
- "Medium - should have mitigation plan"
- "Low - acceptable risk"
```

## Output

After decision-making:

```markdown
## Decisions Complete

**Decisions Made:** [N]
**Status:** [X] accepted, [Y] provisional
**Risks Identified:** [Z]

### Decisions Summary
| ID | Decision | Status | Evidence Count |
|----|----------|--------|----------------|
| DEC-scope-smb-first | Target SMB first | accepted | 4 |
| DEC-pricing-freemium | Use freemium model | provisional | 3 |
| ... | ... | ... | ... |

### Risks Created
| ID | Risk | Severity | Linked Decision |
|----|------|----------|-----------------|
| RISK-market-smb-churn | SMB churn rate | medium | DEC-scope-smb-first |
| ... | ... | ... | ... |

### Quality Gate Status
- Decision gate: ✓ All decisions cite ≥2 evidence
- Risk gate: ✓ All risks have mitigations

### Next Step
Run `/ledger-spec` to generate constrained PRD and architecture.
```

## References

- [references/decision-ledger-schema.md](references/decision-ledger-schema.md) - DECISIONS.yaml schema
- [references/risk-ledger-schema.md](references/risk-ledger-schema.md) - RISKS.yaml schema

Files in this skill

  • SKILL.md5.9 KB
  • references/decision-ledger-schema.md7 KB
  • references/risk-ledger-schema.md7.9 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.