| name | token-budget-advisor-pm |
| description | Offers product managers and teams an informed choice about response depth and token consumption before answering. Specialized for PM workflows: roadmaps, PRDs, market analysis, feature specs, team coordination. Use when the user explicitly wants to control response length, depth, or token budget for PM tasks. TRIGGER when: "token budget", "response length", "short version", "brief", "detailed", "exhaustive", "quick answer", "full analysis", "executive summary vs deep dive", "I need a 5-minute read", "give me the full thing", or clear variants where user is controlling answer size/depth for PM decisions. DO NOT TRIGGER when: user already specified depth level this session (maintain it), or request is a simple fact lookup. |
| origin | ECC + PM specialization |
| tier | professional |
| version | 1 |
Token Budget Advisor for Product Managers (TBA-PM)
Intercept the response flow to offer product managers a choice about response depth before answering PM-specific questions.
When to Use
Use TBA-PM when the user wants to:
- Control how long or detailed a response is
- Get an executive summary vs. deep dive
- Choose between "quick answer" and "full analysis"
- Decide on response depth upfront to save time
- Balance depth against attention/time available
- See options before Claude commits tokens
PM-Specific Triggers:
- "Give me the PRD in 2 pages, not 10"
- "Quick roadmap outline first, then detailed"
- "Executive summary only"
- "Deep dive on this market analysis"
- "I need this in 5 minutes"
- "Full competitive analysis, no shortcuts"
- "Short version of the GTM strategy"
- "Detailed breakdown by quarter"
Do NOT trigger when:
- User already set depth level this session (maintain it silently)
- Request is a simple factual lookup ("What's Salla's GTM?" → just answer)
- User says "token" referring to auth/API tokens, not response size
PM Response Complexity Classification
Product management questions vary significantly in scope. Use this classification to estimate token ranges:
Complexity Levels
| Complexity | Type | Examples | Typical Multiplier |
|---|
| Simple | Single-answer factual | "When does Vision 2030 end?" | 3× – 8× |
| Medium | Process explanation | "How do we structure the roadmap?" | 8× – 20× |
| Medium-High | Structured planning | "Draft PRD outline for [feature]" | 10× – 25× |
| Complex | Multi-part strategy | "Competitive analysis + GTM strategy" | 15× – 40× |
| Creative/Strategic | Vision, narratives, frameworks | "Build 18-month roadmap with KPIs" | 15× – 45× |
How Token Budget Advisor Works for PMs
Step 1 — Estimate Input Tokens
Use heuristic calibration:
- Prose (roadmaps, briefs, market analysis):
words × 1.3
- Code-heavy or mixed:
characters / 4
- PM documents (PRDs, competitive docs):
words × 1.3 (mostly prose)
Example:
User asks: "Create a 6-month product roadmap for our B2B SaaS with KPIs"
Estimated input: ~150 words = ~195 tokens
Step 2 — Classify Complexity & Estimate Response Window
Apply PM-specific complexity classification:
Question: "Create a 6-month product roadmap for our B2B SaaS with KPIs"
Complexity: Medium-High (structured planning)
Multiplier range: 10× – 25×
Input tokens: ~195
Response window:
Min: 195 × 10 = ~1,950 tokens
Max: 195 × 25 = ~4,875 tokens
Step 3 — Present Depth Options for PMs
Show this before answering (for PM-specific workflows):
Analyzing your PM request...
Input: ~[N] tokens | Type: Product Planning | Complexity: [level] | Language: [lang]
Choose your roadmap/analysis depth:
[1] Executive (25%) -> ~[tokens] Key takeaways only, headlines
[2] Balanced (50%) -> ~[tokens] Summary + structure + examples
[3] Detailed (75%) -> ~[tokens] Full roadmap with KPIs, risks, timelines
[4] Exhaustive (100%) -> ~[tokens] Complete with contingencies, all details
Which level? (1-4 or say "executive summary", "balanced", "detailed", "full deep dive")
For roadmaps: 1=outline only | 2=quarterly breakdown | 3=monthly + KPIs | 4=daily tasks + dependencies
For analysis: 1=key findings | 2=findings + context | 3=analysis + recommendations | 4=everything
Precision: heuristic estimate ~85-90% accuracy (±15%).
Depth Levels for PM Tasks
Level 1: Executive (25% of window)
Use for: Quick decisions, leadership briefs, time-constrained PMs
Target length: 1-2 paragraphs
Include:
- Key takeaway / recommendation
- Top 1-2 supporting points
- Decision required
Omit:
- Context, details, edge cases
- Multiple alternatives
- Supporting analysis
- Timelines, risk analysis
Example (Roadmap):
RECOMMENDATION: Launch Feature A in Q1, Feature B in Q2.
Key drivers: Feature A addresses top customer request (40% of support
tickets). Feature B unlocks next revenue tier. No blocking dependencies.
Decision: Proceed with this sequence?
Level 2: Balanced (50% of window)
Use for: Most PM work, standard roadmaps, team coordination
Target length: 2-4 paragraphs or 1 section per topic
Include:
- Full recommendation
- Context + reasoning
- 1-2 key examples
- Top risks or dependencies
- Necessary timelines
Omit:
- Deep analysis, edge cases
- Exhaustive alternatives
- Detailed contingencies
- Minute-level planning
Example (Roadmap):
ROADMAP OUTLINE
Q1: Feature A (Core)
→ Addresses top customer request (40% of support tickets)
→ 6-week build + 2-week testing
→ Owner: [Team]
→ Success metric: Reduce support tickets by 30%
Risk: Dependent on API integration (owned by [Team])
Q2: Feature B + Scaling
→ Unlocks enterprise tier (potential $500K ARR)
→ 8 weeks + 2 weeks optimization
→ Owner: [Team]
Risk: Requires new infrastructure
Q3: Regional Expansion (GCC focus)
→ Market opportunity: $2.5B TAM
→ Localization + partnerships (8 weeks)
→ Owner: [Team]
KPIs by quarter: [defined]
Level 3: Detailed (75% of window)
Use for: Deep strategic work, complex competitive analysis, stakeholder planning
Target length: Multiple sections, structured breakdown
Include:
- Complete strategy with options
- Pros/cons for each option
- Detailed KPIs + success metrics
- Timeline: quarterly + monthly
- Dependencies and risks
- Resource requirements
- Contingency plans
Omit:
- Exhaustive edge case analysis
- Every possible alternative
- Daily-level task breakdown
Example (Roadmap):
PRODUCT ROADMAP - 6 MONTHS (Detailed)
STRATEGY
Pillars: Core stability, customer traction, market expansion
North Star: $2M ARR by end of Q2
Risk tolerance: Medium (validate before scaling)
Q1: STABILITY & CUSTOMER RETENTION (Jan-Mar)
Feature A: Customer request automation
→ Business case: 40% of support tickets
→ Effort: 6 weeks dev + 2 weeks test
→ Owner: Backend team (4 people)
→ Success metric: Support tickets down 30%
→ Risk: Third-party API integration (2-week dependency)
→ Contingency: In-house solution (6 extra weeks)
Q2: MONETIZATION (Apr-Jun)
Feature B: Enterprise tier
→ Business case: $500K+ ARR opportunity
→ Effort: 8 weeks + 2 weeks optimization
→ Owner: Full stack + DevOps
→ Success metric: 5 enterprise customers, $50K MRR
→ Risk: Custom integrations (timeline variable)
→ Contingency: Phased rollout (single integration first)
Scaling: Infrastructure optimization
→ Database optimization (2 weeks)
→ Caching layer (3 weeks)
→ Load testing (1 week)
→ Owner: Platform team
Q3: EXPANSION (Jul-Sep)
Market: GCC tech ecosystem
→ TAM: $2.5B (B2B SaaS segment)
→ Effort: Localization (4 weeks) + partnerships (4 weeks)
→ Owner: Product + BD
→ Success metric: 10 customers, $100K ARR in GCC
→ Risk: Payment integrations, regulatory compliance
→ Contingency: UAE-first, then KSA rollout
KPIs & SUCCESS METRICS
Q1: Support tickets -30%, NPS +5 points
Q2: $50K MRR, 5 enterprise customers, system uptime 99.95%
Q3: GCC revenue $100K ARR, integration partnerships: 3
DEPENDENCIES & RISKS
[Detailed breakdown per quarter]
RESOURCE REQUIREMENTS
Q1: 4 backend, 2 frontend, 1 QA
Q2: 3 backend, 3 frontend, 1 DevOps, 1 QA, 1 Product
Q3: 2 backend, 2 frontend, 1 Product, 1 BD, 1 Country Manager
CONTINGENCY PLANS
[If dependency slips, if competitor launches, if market changes...]
Level 4: Exhaustive (100% of window)
Use for: Strategic bets, complete market research, major product directions, board updates
Target length: No limits, everything included
Include:
- All alternatives with analysis
- Detailed competitive landscape
- Market research data (sourced)
- Quarterly + monthly + weekly breakdown
- Full risk register
- All contingencies
- Detailed resource allocation
- Success metrics + leading indicators
- Change management plan
- Stakeholder communication strategy
No limits on depth.
PM-Specific Depth Shortcuts
If user signals a level, respond at that level immediately without asking:
| What They Say | Level | In PM Context |
|---|
| "1" / "executive summary" / "short version" | 25% | Headlines, key decision |
| "2" / "balanced" / "standard roadmap" | 50% | Quarterly + KPIs |
| "3" / "detailed" / "full roadmap" / "deep dive" | 75% | Monthly detail, all risks |
| "4" / "exhaustive" / "everything" / "board-level" | 100% | No limits, full analysis |
If user set a level earlier in session, maintain it silently unless they change it.
PM Task Type Mapping
Different PM tasks benefit from different default depths:
Quick Decisions
- Default: Level 2 (Balanced)
- Examples: "Should we pivot to GCC?" "Launch feature on schedule?"
- Reasoning: PMs need context but must decide fast
Roadmap Planning
- Default: Level 3 (Detailed)
- Examples: "Create 6-month roadmap" "Plan Q1 priorities"
- Reasoning: Roadmaps require structure, KPIs, dependencies
Market/Competitive Analysis
- Default: Level 3-4 (Detailed to Exhaustive)
- Examples: "Competitive analysis" "TAM estimation" "Market entry strategy"
- Reasoning: Analysis only valuable when comprehensive
PRD/Specification
- Default: Level 3 (Detailed)
- Examples: "Draft PRD for feature" "Write spec for API"
- Reasoning: Needs to be actionable for engineering, but not exhaustive
Team Communication
- Default: Level 1-2 (Executive to Balanced)
- Examples: "Explain feature to CEO" "Brief the team on roadmap"
- Reasoning: Communication needs brevity but context
Strategic Planning
- Default: Level 4 (Exhaustive)
- Examples: "Build 18-month vision" "Strategic planning for next year"
- Reasoning: Strategy requires completeness
How to Respond at Each Level
Responding at Level 1 (Executive/25%)
Keep to 2-4 sentences max.
KEY RECOMMENDATION: [Single clear recommendation]
WHY: [1-2 supporting points]
Next step: [Decision or action]
Responding at Level 2 (Balanced/50%)
SUMMARY: [Clear overview]
KEY POINTS:
• [Point 1]
• [Point 2]
EXAMPLE: [One concrete example]
RISKS: [1-2 top risks]
TIMELINE: [When this matters]
Responding at Level 3 (Detailed/75%)
STRATEGY: [Full approach]
BREAKDOWN:
Phase 1: [Details]
Phase 2: [Details]
Phase 3: [Details]
PROS & CONS:
✅ Advantages
❌ Risks
KPIs & SUCCESS:
[Defined metrics]
DEPENDENCIES:
[What needs to happen first]
TIMELINE:
[Quarterly/monthly breakdown]
RESOURCES NEEDED:
[Team, budget, tools]
Responding at Level 4 (Exhaustive/100%)
[No structure constraints. Everything included.]
Complete analysis with:
- All alternatives
- Detailed comparisons
- Full market research (sourced)
- Quarterly + monthly + weekly timelines
- Complete risk register
- All contingencies
- Resource allocation
- Success metrics + leading indicators
- Change management
- Stakeholder communication strategy
Precision & Limitations
This skill uses heuristic estimation, not real tokenization:
- Accuracy: ~85-90%
- Variance: ±15%
- Always show the disclaimer
Real accuracy depends on:
- Content type (code estimates vary more)
- Model's instruction handling
- Actual token efficiency
PM-Specific Use Cases
Case 1: Quick Market Entry Decision
User: "Should we enter the GCC market? Quick answer only."
TBA Response:
Input: ~250 tokens | Type: Strategic Decision | Complexity: Complex | Language: English
Since you asked for "quick answer," I'll respond at Level 1 (Executive).
[Response at 25% depth: 2-3 sentences max with key decision]
Case 2: 6-Month Roadmap Planning
User: "Create a detailed 6-month roadmap with KPIs and dependencies."
TBA Response:
Input: ~180 tokens | Type: Roadmap | Complexity: Medium-High | Language: English
This typically needs structure and detail. I'll present options:
[1] Outline only (key themes, no detail)
[2] Quarterly breakdown (standard roadmap)
[3] Monthly + KPIs + risks (detailed planning)
[4] Everything (daily tasks, contingencies, resourcing)
Which level? (or I can default to Level 3 for roadmaps)
Case 3: Competitive Analysis
User: "Analyze our top 5 competitors in the GCC B2B SaaS market."
TBA Response:
Input: ~200 tokens | Type: Competitive Analysis | Complexity: Complex | Language: English
Competitive analysis is valuable when comprehensive. I'll do Level 3 (Detailed):
[Full competitor breakdown with positioning, strengths, weaknesses, gaps]
Want the Level 4 exhaustive version? (adds market sizing, pricing analysis,
strategic recommendations)
Case 4: PRD Writing
User: "Draft a PRD for our new customer analytics feature. I need it actionable
for engineering but not overwhelming."
TBA Response:
Input: ~220 tokens | Type: PRD | Complexity: Medium-High | Language: English
PRDs work best at Level 3 (Detailed). I'll structure it for engineering clarity:
[Complete PRD with problem, solution, acceptance criteria, dependencies,
but not exhaustive edge cases]
Want to add Level 4 (exhaustive alternative approaches, all edge cases)?
When TBA-PM Should Trigger
✅ TRIGGER
- "Give me the short version"
- "Detailed roadmap with all KPIs"
- "I need an executive summary"
- "Full competitive analysis"
- "Quick market sizing, not exhaustive"
- "Deep dive on this feature"
- "I want the complete PRD"
- "Just the key takeaway"
❌ DON'T TRIGGER
- "What is Salla's main product?" (simple fact)
- "When does Vision 2030 end?" (lookup, not strategy)
- "Is the GCC market growing?" (yes/no)
- User already said "detailed" or "quick" earlier this session (maintain it)
Quality Gate Checklist
Before presenting depth options:
✅ READINESS
- [ ] User is actually controlling response depth (not just asking a question)
- [ ] This is a PM task (roadmap, analysis, strategy, planning)
- [ ] User said something triggering TBA (token, depth, brief, exhaustive, etc.)
- [ ] User hasn't already set a level this session
✅ ESTIMATION
- [ ] Input tokens estimated correctly (~±15%)
- [ ] Complexity classification appropriate for PM task
- [ ] Response window realistic for the task
- [ ] Token estimates shown clearly
✅ PRESENTATION
- [ ] Depth options are PM-relevant
- [ ] Examples given for each level
- [ ] Expected length and content clear
- [ ] Precision disclaimer included
Summary for PM Workflows
TBA-PM helps PMs:
- ✅ Control response depth upfront (save time vs. oversupply)
- ✅ Get executive summaries when needed
- ✅ Get detailed planning when needed
- ✅ Make informed depth choices
- ✅ Manage token/attention budget
How it works:
- User signals depth preference (or TBA offers options)
- Estimate input tokens + classify complexity
- Present 4 depth levels with token estimates
- User chooses (or TBA defaults for PM task type)
- Respond at chosen level, maintain for session
PM Default Depths:
- Quick decisions: Level 2 (Balanced)
- Roadmaps: Level 3 (Detailed)
- Analysis/research: Level 3-4 (Detailed to Exhaustive)
- Team communication: Level 1-2 (Executive to Balanced)
- Strategic planning: Level 4 (Exhaustive)
