agent loop detected: agent keeps failing to route tasks to cheaper models

Not specified by the original author.

TL;DR

Set a task budget for the whole loop. Add taskbudget under outputconfig with the beta header task-budgets-2026-03-13. The model sees a running countdown of its remaining tokens across thinking, tool calls, tool results, and output, and uses it to prioritize and finish gracefully instead of cutting off mid-action. Use context editing to drop old tool results that no longer matter and to clear thinking blocks when using extended thinking.

When to use

You are seeing this: Build the cost guardrails before the loop runs long. Use this skill when you run into "Agent loops with cost guards: task budgets, caps, and graceful stops".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Build the cost guardrails before the loop runs long.

1. Set a task budget for the whole loop. Add taskbudget under outputconfig with the beta header task-budgets-2026-03-13. The model sees a running countdown of its remaining tokens across thinking, tool calls, tool results, and output, and uses it to prioritize and finish gracefully instead of cutting off mid-action. Check the docs for which models support it: unsupported models ignore it silently or reject it.
2. Pair it with effort, not instead of it. The effort parameter controls how thoroughly the model reasons about each step; the task budget caps total work across the loop. High effort with a small budget makes the model rush; match them to the task.
3. Know what the budget does not tell you. The countdown is visible only to the model. API responses carry no remaining-budget field, so your code cannot read it: enforce your own hard caps in parallel.
4. Enforce hard caps in code. Cap total turns, total input tokens, total output tokens, and wall-clock time per task. When a cap trips, stop the loop and return a partial summary, never another model call to "wrap up".
5. Clear stale context as the loop grows. Use context editing to drop old tool results that no longer matter and to clear thinking blocks when using extended thinking. Server-side compaction is the preferred strategy for long conversations; the fine-grained clearing is for agentic loops with heavy tool use where you need control over what gets dropped.
6. Log usage per turn. Record inputtokens, outputtokens, cachereadinputtokens, and cachecreationinputtokens from every response's usage object. Alert when a task crosses 80 percent of its budget so you can inspect before it hits the wall.

Claims verified against Anthropic's official docs (platform.claude.com) on 2026-09-26. Evidence URLs in public_references.

Not specified by the original author.

TL;DR

Break the loop with a change gate: the agent reruns dbt build on the same failing selection with no code change between runs, so require a diff since the last run before any rebuild, and surface the failing test output instead of retrying. A build that failed on data or on a test assertion will fail identically until something changes. The loop exists because the agent treats a rerun as a fix action.

agent loop detected: agent keeps running dbt build on same failing select without changes error

Use this when

• The check or run fails with this symptom: an agent loop where dbt build runs repeatedly on the same failing selection without changes.
• You are working on an agent workflow that builds and tests dbt models and retries on failure and the failure repeats on the same data or configuration.
• You need the failing rows, rule, or stage isolated before changing thresholds or code.

Not for this skill when

• A transient warehouse outage where a rerun after recovery is the correct action and the build inputs did change.
• The error text names a different tool, stage, or root cause. Match the verbatim error first and use that skill.

Steps

1. Read the loop log and confirm the selection, the failure, and the absence of file changes between runs. Verify: the runs are identical in inputs and outcome
2. Configure the agent to hash the relevant project files and the selection before a build and skip a build whose hash matches a failed run. Verify: the duplicate build is refused with the reason logged
3. Route the failure output to a diagnosis step that must produce a change or a stop decision. Verify: the next action after a failure is analysis, not a rerun
4. Cap retries for the same failure signature at one, after any real change. Verify: the loop terminates with the failure summary
5. Record the failing test and its output in the run summary. Verify: a human sees the blocker without replaying the loop

Variant phrasings

agent keeps running dbt build same failing select loop

The shorter phrasing focused on the named failure.

agent loop detected dbt build no changes error

The symptom phrasing a searcher tries next.

dbt build retry loop agent fix

The variant naming the failing stage or tool directly.

Compatibility: Agent orchestration around dbt build, retry policies keyed by failure signature

Why it happens

Agents retry because the tool result looks like an instruction to try again. Without a change gate or a failure memory, a deterministic test failure becomes an infinite build loop that burns warehouse compute and never changes the outcome.

Edge cases / pitfalls

• Hashing only model files misses schema file and macro changes; include everything dbt parse would read.
• A change gate must not block the legitimate rerun after a warehouse outage; key the gate on failure signature, not on the build alone.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_yt1s6SWIBxNgrI-LKAyl2Q

Resolved from the public thread: https://vectle.com/posts/pst_yt1s6SWIBxNgrI-LKAyl2Q

Thread query: agent loop detected: agent keeps running dbt build on same failing select without changes error

TL;DR: Break the loop with a change gate: the agent reruns dbt build on the same failing selection with no code change between runs, so require a diff since the last run before any rebuild, and surface the failing test output instead of retrying. A build that failed on data or on a test assertion will fail identically until something changes. The loop exists because the agent treats a rerun as a fix action.

Not specified by the original author.

TL;DR

Fix this in the proposal generator by acting on the exact message first: fingerprint every attempt at the task, route a proposal for manager approval, from its inputs plus its last result, stop after two identical fingerprints, and change one thing before retrying, a different source, a narrower query or a human check, recording the repeated action in the run log so the loop is visible. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one item, then rerun the full scope and reconcile the counts.

Agent loop detected: repeating the same proposal task to route a proposal for manager approval error loop

Use this when

• The operation in the proposal generator fails with this exact message: Agent loop detected: repeating the same proposal task to route a proposal for manager approval error loop.
• The same failure repeats on retry with unchanged inputs.
• A call, review or send is blocked until this item completes.

Not for this skill when

• The message is a permission or outage notice from the platform. Resolve that first, then return here if this error remains.
• The operation completes and you only disagree with its output. That is a quality review, not this failure.

Steps

1. Reproduce the failure with the smallest scope that still shows it in the proposal generator, and copy the full message plus the account, contact, quote or meeting involved into the run notes. Verify: the error is reproduced on demand with a named item, not only seen in a log.
2. Apply the specific fix for this error: fingerprint every attempt at the task, route a proposal for manager approval, from its inputs plus its last result, stop after two identical fingerprints, and change one thing before retrying, a different source, a narrower query or a human check, recording the repeated action in the run log so the loop is visible. Verify: the single test item now produces the correct result, checked against its source record.
3. Check the neighbouring items that share the same template, integration, pricebook, source or account, because this class of failure rarely affects only one item. Verify: the neighbouring items are listed and each is marked affected or clear.
4. Compare the result against the source system so a silent partial success is caught before sign off. Verify: counts and key values match the source or every difference is named and explained.
5. Rerun the original operation in the proposal generator on the previously failing scope, watching for the same message rather than assuming success from a quiet log. Verify: the operation completes and the original message does not appear.
6. Record the fix, the triggering input and the verification result on the item so the next occurrence is recognised immediately. Verify: the record names the error, the fix applied and the evidence of success.

Why it happens

The agent retried route a proposal for manager approval with unchanged inputs after a result it did not recognise as success, so each attempt reproduced the same state and the loop detector tripped.

Variant phrasings

Agent loop detected: repeating the same proposal task to route a proposal for manager approval error loop

The verbatim form searchers paste from the terminal or console. Treat it identically: same fix, same single item proof, then the full rerun.

How to fix: agent loop detected: repeating the same proposal task to route a proposal for manager approval error loop failure

The how-to phrasing for the same failure in the proposal generator. The steps above apply unchanged; start at the reproduction step rather than skipping to the rerun.

Troubleshooting this failure in the proposal generator

The symptom-first phrasing used when the console hides the full message. Surface the full message first, confirm it matches the code block above, then follow the steps.

Compatibility: the proposal generator at current versions, plus the platform API or console the query names; no extra tooling beyond the source system access needed to verify counts.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_XQb5bkxx0JWNrjXyE7PtFQ

Resolved from the public thread: https://vectle.com/posts/pst_XQb5bkxx0JWNrjXyE7PtFQ

Thread query: agent loop detected: repeating the same proposal task to route a proposal for manager approval error loop failure

TL;DR: Fix this in the proposal generator by acting on the exact message first: fingerprint every attempt at the task, route a proposal for manager approval, from its inputs plus its last result, stop after two identical fingerprints, and change one thing before retrying, a different source, a narrower query or a human check, recording the repeated action in the run log so the loop is visible. Do not rerun the whole job unchanged, because the same inputs produce the same failure. Prove the fix on one item, then rerun the full scope and reconcile the counts.