Exam Cheat-Sheet — Rules, Values & Anti-Patterns
CCA Foundations course · Page 8 of 8 (final) · ← Back to all courses · This is the rapid-recall sheet. The 3 Universal Rules below answer roughly 80% of questions.
The 3 Universal Rules (answer ~80% of questions)
- Deterministic over probabilistic when stakes are high → use hooks / schemas / forced
tool_choice, not prompt instructions, for financial / compliance / security. - Root cause over symptom patch → fix tool descriptions (not a routing layer); fix coordinator decomposition (not more subagents); fix the schema (not retry logic).
- Simplest correct intervention → don't over-engineer; don't add complexity on top of broken architecture; prefer the minimal fix that addresses the actual cause.
Exam-Day Decision Shortcuts
| When you see… | Answer |
|---|---|
| "financial / compliance / security" | Hooks — PreToolUse to block, PostToolUse to normalise |
| "tool misrouting" | Fix tool descriptions (not few-shot, not tool_choice) |
| "narrow results from multi-agent" | Fix coordinator decomposition (not subagents) |
| "CI pipeline hangs" | Missing -p flag |
| "team standards location" | .claude/CLAUDE.md (always) |
| "cross-directory conventions" | .claude/rules/ with glob frontmatter |
| "parallel agents" | Multiple Task calls in ONE response |
| "context filling up with exact values" | Case-facts block at TOP of every prompt |
| "conflicting statistics" | Annotate both with source + date; never pick or average |
| "explicit human request" | Escalate immediately, no resolution attempt |
| "absent data causing retry loop" | Nullable fields + remove from required |
| "self-review" | Independent instance with zero prior context |
| "18+ tools per agent" | Split into scoped agents (4–5 tools each) |
| "timeout returns empty success" | Silent suppression — return structured error with isError |
| "multiple tools, different formats" | PostToolUse hook to normalise before Claude sees |
| "agent crash recovery" | Periodic JSON manifest export + inject on resume |
Critical Values (instant recall)
| Pass score | 720 / 1000 |
| Total questions | 60 MCQ |
| Duration | 120 minutes |
| Scenarios on exam | 4 of 6 |
| Results SLA | up to 10 business days |
| Exam cost | free for partners / $99 |
stop_reason → continue |
"tool_use" |
stop_reason → stop |
"end_turn" |
stop_reason → truncated |
"max_tokens" |
| Batch API savings | 50% vs synchronous |
| Batch max window | 24 hours (no SLA) |
| Batch multi-turn tools | Not supported |
| Max tools per agent | 4–5 optimal |
| Too-many-tools threshold | 18+ causes degradation |
@import max nesting |
5 levels |
| CI non-interactive flag | -p / --print |
| Attention-dilution threshold | 14+ files in one pass |
File Locations
| Path | Meaning |
|---|---|
~/.claude/CLAUDE.md |
Personal prefs — NOT VCS |
.claude/CLAUDE.md |
Team standards — VCS |
src/api/CLAUDE.md |
Directory-specific — VCS |
.claude/rules/*.md |
Cross-dir glob rules — VCS |
.claude/commands/ |
Team slash commands — VCS |
~/.claude/commands/ |
Personal commands — NOT VCS |
.claude/skills/ |
Team skills — VCS |
~/.claude/skills/ |
Personal skills — NOT VCS |
.mcp.json |
Team MCP servers — VCS |
~/.claude.json |
Personal MCP — NOT VCS |
The 22 Anti-Patterns = the distractor bank
Every wrong answer on the exam is built from an anti-pattern. The fastest way to score is to recognise the anti-pattern and eliminate that option instantly. They're embedded throughout pages 2–7 (each domain's "wrong answer" callouts and the scenario anti-pattern lists) — review those until each one is obvious on sight.
Suggested study plan (~20–30 hrs over 7–10 days)
Phase 1 — Core concepts (60–70%): Day 1–2 Domain 1 · Day 3 Domain 2 · Day 4 Domain 3 · Day 5 Domain 4 · Day 6 Domain 5. Phase 2 — Exam-day essentials (20–30%): Day 7 anti-patterns drill + 6-scenario walkthrough · Day 7–8 critical values & file locations (instant recall). Phase 3 — Practice & gap closure: Day 8–9 official + third-party practice exams · Day 9 review every wrong answer, drill weak domains · Day 10 final recall check (22 anti-patterns + critical values + 3 rules).
🎯 You've reached the end of the course. Mark this page complete to hit 100% — then run timed practice exams and review any domain that's still shaky. Good luck!