GXCafe/ai-agent-failure-logs

Autonomous AI Agent Failure Logs

What this is

Failure records from an autonomous AI agent organization running on local LLMs. Every record is a real failure that occurred in production, not a synthetic example.

• Records: 521
• Period covered: 2026-06-24 to 2026-08-06 (43 days)
• Failure classes: 5

Why these records exist

Monitoring told us the runs succeeded. Utilisation sat at 97-100% for a full day. Deliverables shipped in that day: zero.

The failures below are the reason. Every one of them returned without an error, with plausible length and shape — a refusal recorded as a deliverable, a hedge that passed a length check, an answer to a different question, fluent text in the wrong language, invented field names. None of it is visible to a tracing tool.

The checker we built to reject these is published separately as honto-contract (MIT). This dataset is the evidence it was built from, and the regression cases it is tested against.

Who it is for

Teams running LLM agents in production who want to know what actually breaks, and engineers building guardrails who need real failure cases to test against.

What is inside

One JSON object per line. Fields:

Field	Description
source	Which subsystem recorded it (task_execution or local_inference)
at	Timestamp (UTC)
stage	The kind of work that failed
failure_class	The failure type (see below)
violations	The exact contract clauses that were breached, as recorded
violation_types	Stable machine-readable names for those breaches (see below)
detail	What the failure was, as recorded at the time
duration_sec / duration_ms	How long it ran before failing
model	The model in use, where recorded

What broke, specifically

failure_class is coarse: most production failures land in contract_violation. The useful part is violation_types — the specific clause that was breached. These names are stable across releases and across languages, so they can be counted.

Type	Count	What it means
too_short	143	Output shorter than the contract requires — often empty after the model spent its budget on reasoning
missing_required	80	A required marker was absent (a verdict line, a heading, a field)
other	26	Recorded, not yet classified
chinese_in_japanese	13	Chinese-only characters appeared in Japanese output
unearned_claim	11	Customers, adoption numbers, or superlatives the company does not have
broken_language_mix	5	Japanese and English fused into something neither language
not_japanese	1	Japanese requested, not a single kana returned
reasoning_leaked	1	Internal reasoning leaked into the deliverable

Records without a type: 307 — process failures and timeouts, plus everything recorded before 2026-08-05, when the reasons were not yet written down. A checker that says only "the contract was not met" teaches nobody anything; that gap is why the reasons are recorded now.

Failure classes in this release

Class	Count	Meaning
contract_violation	430	Output did not satisfy the declared contract (length, required marker, format, language)
scheduler_starvation	26	Work was pending but the scheduler could not pick it up
language_corruption	26	Output mixed languages or leaked characters from another language
process_error	24	The inference process failed to start or exited abnormally
timeout	15	Inference exceeded the time limit and was killed

Limitations

• This is a small first release. The instrumentation that produces these records was added on 2026-08-03, so the window is 43 days. The organization itself has been running longer, but those earlier runs were not recorded in this form.
• All records come from one organization, so they may not generalize.
• Text fields are in Japanese where the system logged them in Japanese.
• This dataset makes no claims about adoption, customers, or revenue. There are none.

How it is produced

scripts/build-failure-dataset.js reads the raw operational logs and emits this file. Counts and dates in this card are computed by that script, not written by hand.

Generated 2026-08-06 10:44:26 JST.