Execution And Reliability
Execution flow, blocking behavior, atomicity, and reliability under agent orchestration.
Failure modes: 8 active | 🔴 4 critical · 🟠 3 high · 🟡 1 medium
| File | Severity | Summary |
|---|---|---|
| 10-critical-interactivity.md | 🔴 Critical | Agents run in non-interactive environments |
| 11-critical-timeouts.md | 🔴 Critical | Agents have finite time budgets per tool call |
| 12-critical-idempotency.md | 🔴 Critical | Agents retry on failure |
| 13-critical-partial-failure.md | 🔴 Critical | Multi-step commands can fail mid-execution, leaving the system in an unknown intermediate state |
| 14-high-arg-validation.md | 🟠 High | Many CLI tools begin executing — creating files, sending requests, modifying state — before validating all their argu... |
| 15-high-race-conditions.md | 🟠 High | Agents may invoke multiple tool calls in parallel |
| 16-high-signal-handling.md | 🟠 High | Agents enforce time budgets by killing processes (SIGTERM, then SIGKILL) |
| 17-medium-child-process-leakage.md | 🟡 Medium | CLI tools frequently spawn background processes — log forwarders, file watchers, health monitors, connection pools |
Detailed Metrics
| Challenge | Severity | Frequency | Detectability | Token Spend | Time | Context |
|---|---|---|---|---|---|---|
| §10 | 🔴 Critical | Common | Hard | High | Critical | Low |
| §11 | 🔴 Critical | Common | Hard | High | Critical | Low |
| §12 | 🔴 Critical | Common | Hard | High | High | Medium |
| §13 | 🔴 Critical | Common | Hard | High | High | Medium |
| §14 | 🟠 High | Common | Medium | Medium | Medium | Low |
| §15 | 🟠 High | Situational | Hard | Medium | Medium | Low |
| §16 | 🟠 High | Situational | Hard | Medium | Medium | Low |
| §17 | 🟡 Medium | Situational | Hard | Low | Low | Low |