Agent Budget Exhausted Halfway Through the Task
Your agent burns through its token or cost budget before finishing, leaving work incomplete. Here's how to diagnose spending and resume safely.
#Agents
Articles tagged with #Agents
Restored Agent Checkpoint Is Corrupted
Your agent resumes from a checkpoint but the state is garbled, missing fields, or internally inconsistent. Here's how to detect corruption and recover safely.
Cost Tracking Misses Sub-Agent Usage
Your pipeline's cost report shows $2 but the invoice says $18. Sub-agent token usage is not attributed to the parent run. Here's how to close the gap.
Cycle in Agent Call Graph Goes Undetected
Agents call each other in a loop that never terminates because the orchestrator has no cycle detection. Here's how to find cycles before they run forever.
Agent Handoff Loses Context Between Steps
Your multi-agent pipeline drops critical context at handoff boundaries. Learn to identify the gap and wire durable state in under an hour.
Agent Orchestrator Deadlocks Waiting on Each Other
Two agents block forever waiting for each other's output — a classic deadlock in async pipelines. Detect the cycle and break it in minutes.
Agent Output Leaks Secrets Into Downstream Logs
API keys, tokens, and passwords in agent output get written to logs and traces. Here's how to detect secret leakage and scrub it before it reaches storage.
Agent Output Not Machine-Parseable Downstream
Your agent wraps JSON in markdown or adds prose commentary, breaking the downstream parser. Here's how to enforce structured output reliably.
Agent Skipped a Pre-flight Check It Was Supposed to Run
Your agent proceeded without running required pre-flight checks, causing avoidable failures downstream. Here's how to enforce mandatory checks before execution.
Promotion Criteria Too Loose — Bad Output Slips Through
Your agent pipeline promotes flawed outputs because gate criteria are too broad or easy to satisfy. Here's how to tighten gates without over-blocking.
Prompt Templates Drift Between Agent Versions
Different versions of your agent use subtly different prompts, causing inconsistent output across the fleet. Here's how to version prompts and detect drift early.
One Agent's Rate Limit Cascades Into a Chain Failure
A single rate-limited agent stalls the entire pipeline as upstream agents queue and timeout. Learn to isolate rate limits and prevent cascade failures.
Flaky Tool Triggers an Agent Retry Storm
One unreliable tool call causes your agent to retry hundreds of times, exhausting budget and rate limits. Here's how to add backoff and circuit-breaking.
Shared Memory Corrupted by Overlapping Agent Writes
Two agents write to the same shared memory store simultaneously, producing garbled or inconsistent state. Here's how to detect and prevent overlapping writes.
Agent Skipped a Required Validation Step
Your agent bypassed a critical validation gate and pushed bad output downstream. Learn to enforce non-skippable checkpoints in any agent pipeline.
Agent State Desyncs After Restart
After a crash or restart, your agent believes the world is in a state that no longer matches reality. Here's how to detect drift and resync reliably.
Agent's Subprocess Orphaned After Agent Exits
Subprocesses launched by your agent keep running after the agent exits, consuming resources and causing side effects. Here's how to track and clean them up.
Task Routed to the Wrong Agent
Your orchestrator sends tasks to the wrong specialist agent, producing garbage output. Here's how to diagnose routing logic and fix misclassification.
Critical Tool Call Missing From the Agent Trace
Your agent trace shows a result but no corresponding tool call, hiding what actually happened. Here's how to find the gap and make traces complete.
Two Parallel Agents Edit the Same File
Parallel agents overwrite each other's changes, causing lost edits and broken code. Here's how to detect conflicts and enforce file-level locking.
Multi-Agent Handoff Prompts for Claude Code Subagents
12 prompt templates to hand work between Claude Code subagents — research → plan → implement → review → ship — without losing context.