Governance And Risk

Governance and Risk

Overview

This skill implements the Decision Analysis & Resolution (DAR) and Risk Management (RSKM) process areas from the CMMI-based SDLC prescription.

Core principle: Proactive governance prevents costly reactive firefighting. Documentation and risk management are investments that pay 3-10x returns by avoiding crisis mode.

Critical distinction:

• Reactive: Handle problems when they occur (expensive, stressful, compounding)
• Proactive: Identify and mitigate problems before they occur (cheap, controlled, preventive)

Reference: See docs/sdlc-prescription-cmmi-levels-2-4.md Sections 3.4.1 (DAR) and 3.4.2 (RSKM) for complete policy.

When to Use

Use this skill when:

• Making architectural or technical decisions without ADRs
• Hearing "it's obvious" or "everyone agrees" (groupthink red flag)
• Skipping risk identification ("what could go wrong?")
• Accepting risks without mitigation plans
• Deferring to authority without independent analysis (CTO says, tech lead suggests)
• Using sunk cost to justify decisions ("we've already invested...")
• Treating governance as bureaucracy or overhead
• No ongoing risk monitoring ("set and forget")

Do NOT use for:

• Trivial decisions (variable names, code style) → Use coding standards
• Implementation details → Use design-and-build skill
• Security-specific risk analysis → Use ordis-security-architect

Quick Reference

Governance Level Framework

When Practices Are MANDATORY

Level 2 Baseline (All Projects):

• ADRs for high-risk decisions (vendor selection, framework choice, data storage)
• Risk identification with basic register
• Mitigation plans for high-probability or high-impact risks

Level 3 Organizational Standard:

• ADRs for all architectural decisions (not just high-risk)
• Alternatives analysis with decision criteria
• Risk register with probability/impact classification
• Scheduled risk reviews (not set-and-forget)
• Independent analysis before authority/consensus input

Level 4 Quantitative:

• Statistical risk models
• Quantitative decision criteria
• Process performance baselines for decision quality

When Practices Are OPTIONAL

Level 1 or Low-Risk Projects:

• Internal prototypes (< 2 week lifespan)
• Single-developer projects with no audit requirements
• Throwaway code (spikes, experiments)

CRITICAL: "Low-risk" is often optimism bias. Verify with risk assessment before declaring optional.

Anti-Patterns and Rationalizations

"It's Obvious"

Detection: "Everyone agrees", "clear choice", "no brainer"

Why it's tempting: Saves time, reduces documentation burden, team aligned

Why it fails: Today's "obvious" is tomorrow's mysterious. Future maintainers lack context, assumptions not validated, alternatives not considered

Counter:

• Level 3 requirement: Document even "obvious" decisions
• Context loss timeline: 6 months for team turnover, 3 months for forgotten assumptions
• Question to ask: "If someone joins the team in 6 months, will they know WHY we chose this?"
• Lightweight ADR takes 20 minutes, saves hours of future confusion

Red flags: "We all know", "Obviously", "No need to write it down"

"Low-Risk Project"

Detection: "Simple project", "Internal only", "We've done this before", "What could go wrong?"

Why it's tempting: Small scope, experienced team, reduces overhead

Why it fails: Scope creep, resource constraints, and timeline slips hit "simple" projects just as often. Optimism bias blinds to risks.

Counter:

• Level 2 requirement: Risk identification for ALL projects
• Common risks for "simple" projects: scope creep (stakeholders add "just one more thing"), resource availability (PTO, competing priorities), data access (permissions, security approvals), timeline slip (integration surprises)
• Reactive firefighting costs 3-10x proactive planning
• 30-minute risk session saves days of crisis mode

Red flags: "What could go wrong?", "It's just...", "Low-risk"

"Authority/CTO Prefers It"

Detection: "CTO met with vendor", "Tech lead suggested", "Management wants"

Why it's tempting: Reduces conflict, speeds decision, aligns with leadership

Why it fails: Authority bias prevents genuine alternatives analysis. Senior stakeholders have blind spots, vendor relationships create bias, title ≠ technical correctness

Counter:

• Level 3 requirement: Independent alternatives analysis BEFORE authority input
• Document decision criteria first (security, cost, integration, vendor stability)
• Evaluate options against criteria WITHOUT authority preference
• Present analysis to authority: "Here's what the data shows, here's your preference, here's my recommendation"
• Authority can override, but must be documented as "decision override based on non-technical factors"

Red flags: "CTO wants", "We should align with leadership", "Don't want to contradict"

"We've Already Invested Time" (Sunk Cost)

Detection: "We've had 2 sales calls", "Demo account set up", "Already started integration"

Why it's tempting: Feels wasteful to "go backwards", momentum toward choice

Why it fails: Sunk cost fallacy - past investment doesn't validate future commitment. Small sunk cost vs large future cost (vendor lock-in, wrong tool).

Counter:

• Name the fallacy: "This is sunk cost fallacy"
• Calculate future cost: "2 sales calls (4 hours sunk) vs 3-year vendor lock-in (hundreds of hours if wrong choice)"
• Reframe: "We invested 4 hours evaluating Option A. Should we invest 2 hours evaluating Options B and C to validate?"
• Past investment gives you evaluation data, not decision commitment

Red flags: "We've already", "Going backwards", "Wasted effort"

"Trust the Vendor" / "99.9% SLA"

Detection: "Established company", "Good reputation", "SLA guarantees uptime"

Why it's tempting: Vendor reputation, SLA promises reduce perceived risk

Why it fails: SLAs are probabilistic, not guarantees. 99.9% = 43 minutes downtime per month. All vendors have outages. Trust ≠ technical mitigation.

Counter:

• Calculate SLA impact: 99.9% uptime = 43 min/month, 8.76 hours/year. Acceptable for your use case?
• Mitigation still required: Circuit breaker, fallback, queueing, graceful degradation
• Vendor reputation reduces probability but doesn't eliminate risk
• Question: "What happens to our users if vendor API is down for 1 hour? Do we have a plan?"

Red flags: "We can trust them", "SLA is good enough", "Reputable company"

"We'll Fix It If It Happens"

Detection: "Handle issues as they come up", "React when needed", "Cross that bridge"

Why it's tempting: Defers work, avoids speculation, focuses on current tasks

Why it fails: Reactive firefighting costs 3-10x proactive mitigation. Incidents occur when you have least capacity to respond (deadlines, weekends, vacations).

Counter:

• Cost math: 1 hour mitigation planning now vs 10 hours firefighting later
• Reactive timing: Incidents don't wait for convenient times - they hit during sprints, before demos, on Friday evenings
• Level 2 requirement: Mitigation plan for high-probability or high-impact risks BEFORE acceptance
• Question: "Do you have 10 hours next week to drop everything and firefight this risk if it materializes?"

Red flags: "We'll handle it", "If it happens", "Cross that bridge when we come to it"

"Risks Haven't Materialized" (Complacency)

Detection: "4 months in, no issues", "Original risks didn't hit", "We're good"

Why it's tempting: Past success validates approach, monitoring feels wasteful

Why it fails: Risks evolve throughout project lifecycle. Absence of risks to-date ≠ absence of future risks. Complacency before late-stage crunch (integration, final testing, deployment).

Counter:

• Lifecycle risk evolution: Early risks (requirements, team ramp-up) vs late risks (integration, tech debt, timeline crunch)
• Month 4 of 6: Integration testing, timeline pressure, technical debt, scope control
• Level 3 requirement: Scheduled risk reviews, not set-and-forget
• New risks emerge, probabilities shift, priorities change

Red flags: "No problems yet", "We're on track", "Monitoring feels like overhead"

"Process Feels Like Bureaucracy"

Detection: "Overhead", "Red tape", "Meetings for meetings' sake", "We want to code"

Why it's tempting: Team wants to deliver, documentation feels unproductive

Why it fails: Lightweight process prevents heavyweight problems. 30 min planning saves hours of firefighting. Process ≠ bureaucracy.

Counter:

• Process vs bureaucracy: Process has ROI (30 min → saves hours). Bureaucracy has no ROI (forms for forms' sake).
• Lightweight governance: 20-min ADR, 30-min risk session, 15-min risk review
• Cost comparison: 30 min process now vs 10+ hours crisis later
• Question: "Would you rather spend 30 minutes planning or 10 hours firefighting next month?"

Red flags: "Bureaucracy", "Overhead", "Red tape", "Slows us down"

"We're Tired / Under Pressure"

Detection: "Just finished major release", "Deadline is tight", "Team exhausted"

Why it's tempting: Exhaustion and deadlines are real, shortcuts feel necessary

Why it fails: Shortcuts under pressure create more pressure later. Technical debt compounds into crisis. Skipping governance creates future exhaustion.

Counter:

• Compound effect: Skipping governance now creates 3x more work later
• Pressure math: 2 hours deadline pressure now vs 10+ hours crisis pressure later
• When you're exhausted is exactly when you need process (prevents mistakes)
• Question: "Will skipping governance make the NEXT deadline easier or harder?"

Red flags: "We're exhausted", "Too busy", "Under pressure", "Just this once"

"We'll Document Later"

Detection: "After we ship", "When we have time", "In the next sprint"

Why it's tempting: Defers effort, focuses on delivery now

Why it fails: "Later" never comes. Context is lost immediately. Future maintainers suffer.

Counter:

• Historical pattern: "Later" has 5% success rate (documented fact)
• Context loss: Starts immediately, complete within 2 weeks
• Requirement: Documentation is part of "done", not optional follow-up
• Question: "When exactly is 'later'? Put it on the calendar now."

Red flags: "Later", "After we ship", "When we have time", "Eventually"

Handling "My Project Is Special" Exceptions

Common exception requests:

• "We're a startup, need to move fast"
• "This is just an MVP/prototype"
• "We'll upgrade to proper governance after product-market fit"
• "Our team is experienced, we don't need process"
• "This project is different because..."

Why it's tempting: Context appears legitimately exceptional, constraints feel unique, team confidence is high

Why it fails: Every team thinks they're special. Startups fail from poor decisions as often as slow delivery. "MVP" and "prototype" often become production. "After product-market fit" never arrives.

Response framework:

1. Acknowledge the constraint: "Startup time pressure IS real, I understand the urgency"

2. Reframe governance as enabler: "Governance prevents the costly mistakes that kill startups - bad vendor choices, unmitigated risks, undocumented decisions that create chaos during scaling"

3. Offer Level 2 as minimum: "Not suggesting heavyweight process. Level 2 = 30-min risk session, ADRs for high-risk decisions only. That's ~2% overhead, not 20%."

4. Make exception criteria explicit:

   • Non-negotiable Level 2 minimum: Security-critical, customer-facing, financially-material, or >3 month projects
   • Level 1 acceptable: Internal prototype <2 weeks, single developer, throwaway code, no audit trail needed
   • Exception must be documented: If declaring Level 1, document why in project README

5. Show the math: "30 minutes planning vs 10+ hours crisis firefighting. Which timeline can your startup afford?"

Red flags indicating exception is rationalization, not legitimate:

• "Just this once" (it's never just once)
• "We'll add governance later" (later has 5% success rate)
• "Trust me, I've done this before" (past success ≠ future guarantee)
• Can't articulate specific criteria for when to upgrade governance

Non-negotiable boundaries:

• Customer-facing code: Minimum Level 2
• Financial transactions: Minimum Level 2
• Security-sensitive (PII, auth, payments): Minimum Level 2

• 3 month timeline: Minimum Level 2

• 5 developers: Minimum Level 3

Example response:

User: "We're a startup MVP, can we skip the ADRs?"

Agent: "I understand the pressure to move fast. Let's find the right balance:

Your MVP is customer-facing with payment processing → Level 2 minimum, not negotiable.

Level 2 = lightweight, not heavyweight:
- ADRs for high-risk decisions only (database, payment gateway, auth provider)
- 30-min risk session (identify risks, plan mitigation for top 3)
- Total overhead: ~2% of project time

This prevents the mistakes that kill MVPs:
- Wrong payment provider → costly migration mid-growth
- No auth security planning → breach before Series A
- Undocumented decisions → chaos when scaling team

ROI: 2 hours planning saves 20+ hours crisis firefighting.

Can we start with risk identification? 30 minutes now."

Reference Sheets

The following reference sheets provide detailed methodologies for specific governance domains. Load them on-demand when needed.

1. Decision Analysis & Resolution (DAR)

When to use: Making architectural decisions, evaluating alternatives, documenting choices

→ See dar-methodology.md

Covers:

• When ADRs are mandatory vs optional
• ADR template and examples
• Decision criteria frameworks
• Alternatives analysis process
• Decision matrix tools
• Authority bias resistance

2. Risk Management (RSKM)

When to use: Identifying risks, assessing probability/impact, planning mitigation, monitoring risks

→ See rskm-methodology.md

Covers:

• Risk identification techniques
• Probability × Impact matrix
• Risk mitigation strategies (avoid, transfer, mitigate, accept)
• Risk register template
• Monitoring and review cadence
• Risk triggers for ad-hoc reviews

3. Templates and Examples

When to use: Need concrete templates for ADRs or risk registers

→ See templates.md

Covers:

• ADR template (lightweight and comprehensive)
• Risk register format
• Decision matrix template
• Real-world examples

4. Level 2→3→4 Scaling

When to use: Understanding appropriate governance rigor for project tier

→ See level-scaling.md

Covers:

• Level 2 baseline practices
• Level 3 organizational standards
• Level 4 quantitative management
• When to escalate or de-escalate rigor

Common Mistakes

Integration with Other Skills

Real-World Impact

Without this skill: Teams experience:

• "Obvious" decisions become mysterious (context loss)
• Authority bias and groupthink (bad decisions)
• Reactive firefighting (3-10x cost)
• No risk mitigation (crisis mode when risks materialize)
• Documentation never happens ("later")

With this skill: Teams achieve:

• Documented decisions with rationale (knowledge retention)
• Independent alternatives analysis (better decisions)
• Proactive risk mitigation (prevent crisis)
• Ongoing risk monitoring (adapt to changing conditions)
• Governance as lightweight process (ROI-positive)

Next Steps

1. Determine project level: Check CLAUDE.md or ask user for CMMI target level (default: Level 3)
2. Identify situation: Use Quick Reference table to find applicable framework
3. Load reference sheet: Read detailed methodology (DAR or RSKM)
4. Enforce requirements: Level 3 requires ADRs for all architectural decisions, risk mitigation for high risks
5. Counter rationalizations: Use anti-pattern catalog to address shortcuts
6. Provide templates: Lightweight ADR or risk register to reduce friction
7. Calculate ROI: Show cost comparison (30 min planning vs 10+ hours firefighting)

Remember: Proactive governance prevents costly reactive firefighting. Documentation and risk management are investments with 3-10x returns.