Proposal Writing

Proposal Writing

When to Use

Use this skill when tasked with writing a proposal, project pitch, RFC, business case, or any persuasive document that recommends a course of action. This is for knowledge work proposals -- recommending strategies, initiatives, process changes, or investments -- not for writing code documentation or API specs.

Common scenarios:

• Proposing a new initiative, project, or program
• Writing an RFC (Request for Comments) for a technical or process change
• Building a business case for investment or resource allocation
• Pitching an approach to solve an identified problem
• Recommending a vendor, tool, or strategy with justification

Principles

1. Lead with the problem, not the solution. The reader must feel the pain before they will accept the cure. Spend proportionate effort defining the problem clearly.
2. Quantify wherever possible. "This will save approximately 40 hours per month" is persuasive. "This will save time" is not.
3. Acknowledge costs and risks honestly. A proposal that ignores downsides loses credibility. Address them head-on and show you have mitigation plans.
4. Write for the decision-maker. The first page should contain everything they need to say yes or no. Details come after.
5. One proposal, one recommendation. If you have multiple options, make a clear recommendation. Do not present a menu and ask the reader to choose.

Output Template

# Proposal: [Title]

**Date:** YYYY-MM-DD
**Author:** [agent name]
**Status:** Draft | Under Review | Approved | Rejected
**Decision Required By:** [date or milestone]

## Problem Statement

### Current Situation
[What is happening now. Be specific and factual.]

### Impact
[Why this matters. Quantify: cost, time lost, risk exposure, missed opportunity.]
| Impact Area | Metric | Current Value | Target Value |
|-------------|--------|---------------|--------------|
| [area] | [metric] | [current] | [desired] |

### Root Cause
[Why the current situation exists. What underlying factors drive it.]

## Proposed Solution

### Overview
[2-3 sentence summary of what is proposed]

### How It Works
[Detailed description of the proposed approach. Break into numbered steps or phases if complex.]

### Why This Approach
[Rationale for choosing this approach over alternatives. Reference specific criteria.]

## Benefits

| Benefit | Description | Estimated Value | Timeline to Realize |
|---------|-------------|-----------------|---------------------|
| [benefit 1] | [description] | [quantified] | [when] |
| [benefit 2] | [description] | [quantified] | [when] |

## Costs and Resources

| Item | Cost | Type | Notes |
|------|------|------|-------|
| [item 1] | [amount] | [one-time/recurring] | [details] |
| [item 2] | [amount] | [one-time/recurring] | [details] |
| **Total** | **[amount]** | | |

## Risks and Mitigations

| Risk | Likelihood | Impact | Mitigation |
|------|-----------|--------|------------|
| [risk 1] | [H/M/L] | [H/M/L] | [what we will do about it] |
| [risk 2] | [H/M/L] | [H/M/L] | [what we will do about it] |

## Alternatives Considered

| Alternative | Pros | Cons | Why Not Chosen |
|------------|------|------|----------------|
| [option A] | [pros] | [cons] | [reason] |
| Do nothing | [pros] | [cons] | [reason] |

## Implementation Plan

### Phase 1: [Name] (Weeks 1-N)
- [ ] [Deliverable 1]
- [ ] [Deliverable 2]

### Phase 2: [Name] (Weeks N-M)
- [ ] [Deliverable 3]
- [ ] [Deliverable 4]

### Timeline

| Phase | Start | End | Key Milestone | Owner |
|-------|-------|-----|---------------|-------|
| Phase 1 | [date] | [date] | [milestone] | [who] |
| Phase 2 | [date] | [date] | [milestone] | [who] |

## Success Criteria

| Criterion | Measurement | Target | Evaluation Date |
|-----------|-------------|--------|-----------------|
| [criterion] | [how measured] | [target value] | [when to check] |

## Decision Requested

[Exactly what you are asking the reader to approve, fund, or authorize. Be specific.]

Procedure

1. Gather Background Materials

TASK_ID="$1"
TOPIC="$2"

# Read the task description
bash /home/shared/scripts/task.sh get "$TASK_ID" | jq -r '.description'

# Find all related documents in workspace
echo "=== Related Documents ==="
rg -l -i "$TOPIC" /home/shared/ ~/workspace/ 2>/dev/null \
  | grep -E '\.(md|txt|csv|json|yaml)$' | head -20

# Check for prior research or analysis that informs this proposal
echo ""
echo "=== Prior Research ==="
find /home/shared/ -name '*research*' -o -name '*analysis*' -o -name '*review*' \
  | grep -i "$TOPIC" 2>/dev/null

# Check for existing proposals on related topics
echo ""
echo "=== Existing Proposals ==="
find /home/shared/ ~/workspace/ -name '*proposal*' -o -name '*rfc*' -o -name '*pitch*' \
  2>/dev/null | head -10

# Read relevant artifacts
echo ""
echo "=== Registered Artifacts ==="
bash /home/shared/scripts/artifact.sh list 2>/dev/null \
  | jq -r '.[] | select(.name | test("'"$TOPIC"'"; "i")) | "\(.name) — \(.path)"' 2>/dev/null

2. Define the Problem

This is the most critical step. A poorly defined problem produces a weak proposal.

# Create a problem definition worksheet
WORK_DIR="/tmp/proposal-${TOPIC}"
mkdir -p "$WORK_DIR"

cat > "$WORK_DIR/problem-definition.md" <<'EOF'
## Problem Definition Worksheet

### What is happening?
[Observable facts, not interpretations]

### Who is affected?
[Specific stakeholders or teams]

### What is the measurable impact?
[Numbers: cost, time, frequency, error rate]

### How long has this been happening?
[Duration and trend: getting worse, stable, improving?]

### What has already been tried?
[Prior attempts and why they did not fully succeed]

### What will happen if we do nothing?
[Projected impact of inaction over 3, 6, 12 months]
EOF

3. Research and Quantify the Impact

# Pull data from available sources to quantify the problem
echo "=== Quantifying Impact ==="

# Search for metrics, numbers, or data related to the problem
rg -n '\b\d+\.?\d*\s*(%|percent|hours|days|dollars|\$|cost|spent|lost|error|fail)' \
  /home/shared/ ~/workspace/ 2>/dev/null | head -20

# Check for any data files with relevant metrics
find /home/shared/ -name '*.csv' -o -name '*.json' \
  | xargs grep -li "$TOPIC" 2>/dev/null

# Look at task history for patterns
bash /home/shared/scripts/task.sh list 2>/dev/null \
  | jq '[.[] | select(.description | test("'"$TOPIC"'"; "i"))] | length' 2>/dev/null

4. Develop the Proposed Solution

# Outline the solution — ensure each element is concrete
cat > "$WORK_DIR/solution-outline.md" <<'EOF'
## Solution Outline

### What will change?
1. [Concrete change 1]
2. [Concrete change 2]
3. [Concrete change 3]

### What stays the same?
- [Unchanged element — important for managing scope expectations]

### What resources are needed?
- People: [who and how much of their time]
- Tools: [what tools or systems]
- Budget: [dollar amount or effort estimate]
- Time: [duration]

### How will we know it worked?
- [Success metric 1 with target value]
- [Success metric 2 with target value]
EOF

5. Assess Costs, Risks, and Alternatives

# Build a cost estimate
python3 <<'PYEOF'
import json

costs = [
    {"item": "Example item 1", "cost": 0, "type": "one-time", "notes": ""},
    {"item": "Example item 2", "cost": 0, "type": "recurring/month", "notes": ""},
]

total_onetime = sum(c["cost"] for c in costs if c["type"] == "one-time")
total_recurring = sum(c["cost"] for c in costs if "recurring" in c["type"])

print(f"One-time costs:  ${total_onetime:,.0f}")
print(f"Recurring costs: ${total_recurring:,.0f}/month")
print(f"First-year total: ${total_onetime + total_recurring * 12:,.0f}")

# Output as table
print("\n| Item | Cost | Type | Notes |")
print("|------|------|------|-------|")
for c in costs:
    print(f"| {c['item']} | ${c['cost']:,.0f} | {c['type']} | {c['notes']} |")
PYEOF

For risks:

# Build a risk register
cat > "$WORK_DIR/risks.md" <<'EOF'
| Risk | Likelihood | Impact | Mitigation |
|------|-----------|--------|------------|
| [What could go wrong] | [H/M/L] | [H/M/L] | [Specific action to reduce] |

Scoring guide:
- H likelihood: >50% chance of occurring
- M likelihood: 10-50% chance
- L likelihood: <10% chance
- H impact: blocks the initiative or causes significant damage
- M impact: delays or degrades the initiative
- L impact: minor inconvenience, easily absorbed
EOF

6. Build the Implementation Timeline

# Create a phased implementation plan
cat > "$WORK_DIR/timeline.md" <<'EOF'
| Phase | Duration | Deliverables | Dependencies | Owner |
|-------|----------|-------------|-------------|-------|
| 1: [Name] | [N weeks] | [what is delivered] | [what must happen first] | [who] |
| 2: [Name] | [N weeks] | [what is delivered] | [Phase 1 complete] | [who] |
| 3: [Name] | [N weeks] | [what is delivered] | [Phase 2 complete] | [who] |

Key milestones:
- Week N: [milestone — a decision point or deliverable]
- Week M: [milestone]
EOF

7. Write the Proposal

PROPOSAL_FILE="/home/shared/proposal-$(date +%Y%m%d)-${TOPIC}.md"

# Assemble the proposal from the working materials
# Read each working file and integrate into the template
cat "$WORK_DIR/problem-definition.md"
cat "$WORK_DIR/solution-outline.md"
cat "$WORK_DIR/risks.md"
cat "$WORK_DIR/timeline.md"

# Write the final proposal (use the Output Template above as structure)
cat > "$PROPOSAL_FILE" <<'PROPOSAL'
# Proposal: [Title]

**Date:** YYYY-MM-DD
**Author:** [agent name]
**Status:** Draft
**Decision Required By:** [date]

## Problem Statement
...

## Proposed Solution
...

## Benefits
...

## Costs and Resources
...

## Risks and Mitigations
...

## Alternatives Considered
...

## Implementation Plan
...

## Success Criteria
...

## Decision Requested
[Exactly what approval or authorization is being sought]
PROPOSAL

echo "Proposal written to: $PROPOSAL_FILE"

8. Self-Review Before Submission

Run these checks on your draft:

PROPOSAL_FILE="/home/shared/proposal-$(date +%Y%m%d)-${TOPIC}.md"

echo "=== Proposal Quality Checks ==="

# Check: Does the problem section have numbers?
echo "--- Quantified impact ---"
rg '\d+' "$PROPOSAL_FILE" | head -10
QUANT_COUNT=$(rg -c '\b\d+\.?\d*\s*(%|hours|days|\$|cost)' "$PROPOSAL_FILE" 2>/dev/null || echo 0)
echo "Quantified claims found: $QUANT_COUNT"
[ "$QUANT_COUNT" -lt 3 ] && echo "WARNING: Proposal needs more quantified evidence"

# Check: Are all required sections present?
echo ""
echo "--- Required sections ---"
for section in "Problem Statement" "Proposed Solution" "Benefits" "Costs" "Risks" "Alternatives" "Implementation" "Success Criteria" "Decision Requested"; do
  if rg -q "$section" "$PROPOSAL_FILE" 2>/dev/null; then
    echo "  OK: $section"
  else
    echo "  MISSING: $section"
  fi
done

# Check: Is the "do nothing" alternative addressed?
echo ""
echo "--- Do-nothing alternative ---"
if rg -qi "do nothing\|status quo\|no action\|inaction" "$PROPOSAL_FILE" 2>/dev/null; then
  echo "  OK: Do-nothing scenario addressed"
else
  echo "  WARNING: Missing do-nothing/status quo alternative"
fi

# Check: Word count (proposals should be thorough but not bloated)
echo ""
echo "--- Length ---"
WC=$(wc -w < "$PROPOSAL_FILE")
echo "  Word count: $WC"
[ "$WC" -lt 300 ] && echo "  WARNING: Proposal may be too thin (< 300 words)"
[ "$WC" -gt 3000 ] && echo "  WARNING: Proposal may be too long (> 3000 words). Consider tightening."

9. Register and Notify

bash /home/shared/scripts/artifact.sh register \
  --name "proposal-${TOPIC}" \
  --type "proposal" \
  --path "$PROPOSAL_FILE" \
  --description "Proposal: ${TOPIC}"

# Notify the requesting agent or reviewer
bash /home/shared/scripts/send-mail.sh \
  --to "$REQUESTING_AGENT" \
  --subject "Proposal ready for review: ${TOPIC}" \
  --body "Draft proposal available at: $PROPOSAL_FILE"

Quality Checklist

• Problem statement is specific, factual, and quantified (not vague or emotional)
• Impact is measured with numbers, not adjectives
• Proposed solution is concrete with specific deliverables
• Benefits are quantified with estimated values and timelines
• Costs include both one-time and recurring, with a first-year total
• At least 3 risks identified, each with a specific mitigation
• "Do nothing" is included as an alternative considered
• Implementation plan has phases, deliverables, and owners
• Success criteria are measurable with target values and evaluation dates
• Decision requested is explicit and specific
• Document leads with the most important information (decision-maker can decide from page 1)
• Report is registered as an artifact in the shared workspace