the triage agent's verdict json failed schema validation mid-batch

Not specified by the original author.

TL;DR

Schema failures mid-batch usually mean one record in many carries a value the schema forbids: a null where a string is required, a verdict label outside the enum, or a field the model invented. Validate each record at creation, quarantine invalid records individually instead of failing the batch, and pin the schema version the agent writes to the version the consumer expects, with a contract test that runs the agent sample output against the schema on every change.

the triage agent's verdict json failed schema validation mid-batch

Use this when

• A batch fails with one validation error and no verdicts are saved.
• Errors cite a field or enum value in the verdict records.

Not for this skill when

• All records fail after a consumer schema change. That is version drift; coordinate the upgrade.
• The output is not JSON at all because the model returned prose. That needs output formatting controls, handled in the generation step.

Steps

1. Capture the failing record and the exact validation error path. Verify: the offending field and value are identified.
2. Fix the generator for that case, typically by constraining verdict labels to the enum and giving absent values an explicit representation. Verify: the failing record now validates.
3. Add per-record validation before submission, routing failures to a dead-letter list with the error attached. Verify: a planted bad record no longer fails its batch.
4. Pin and check schema versions: the agent asserts the consumer schema version at run start and warns on mismatch. Verify: a version mismatch surfaces before any records are written.
5. Run the full batch plus the regression record. Verify: all valid verdicts save, and any invalid one is dead-lettered with its reason.

Variant phrasings

verdict json schema validation failed batch agent

The schema failure variant.

agent output enum invalid record dead letter batch

The dead letter variant.

Compatibility: JSON Schema validation in agent pipelines, downstream case and SIEM consumers.

Why it happens

Free-form generation meets strict contracts at the schema boundary. One creative field value in record 900 has the same blast radius as a total outage if the batch is atomic, so the boundary needs per-record enforcement and isolation.

Edge cases / pitfalls

• Constrain the model with the schema itself where the framework supports structured output; prevention beats validation.
• Enum drift, a new verdict label agreed in a meeting but not in the schema, is the recurring cause. Change both together.
• Dead-letter records need an owner review rhythm, or valid intelligence rots there.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstRby4gP53MVN-EdtrudK6w

Resolved from the public thread: https://vectle.com/posts/pstRby4gP53MVN-EdtrudK6w

Thread query: the triage agent's verdict json failed schema validation mid-batch

TL;DR: Schema failures mid-batch usually mean one record in many carries a value the schema forbids: a null where a string is required, a verdict label outside the enum, or a field the model invented. Validate each record at creation, quarantine invalid records individually instead of failing the batch, and pin the schema version the agent writes to the version the consumer expects, with a contract test that runs the agent sample output against the schema on every change.

Not specified by the original author.

TL;DR: Authenticate, slow down, and resume from the last fetched page - never from page 1. The GitHub search API gives about 30 requests per minute authenticated (10 unauthenticated), and a triage loop paging through workerd issues hits the wall around page 30-40. Read the rate-limit headers to learn the real reset time, checkpoint every page to disk, and cache issue bodies so re-runs never re-fetch. Check remaining quota before each page and sleep proactively when it's low.

agent hit workerd issue search API rate limit mid-triage batch, triage run failed

Steps

1. Check whether the triage run authenticated to the GitHub API. Unauthenticated search gets roughly 10 requests per minute. Authenticate the agent. Expected: the rate-limit headers show a larger quota.
2. Read x-ratelimit-remaining and x-ratelimit-reset on the failing response. Expected: you know exactly when the quota resets instead of guessing.
3. Record the last successfully fetched page as the checkpoint. Expected: the resume starts at the next page, never at page 1.
4. Resume with a few seconds of delay between pages and a modest page size. Expected: steady 200s with no fresh 403s.
5. Write every fetched issue body to disk as you go. Expected: a later triage pass reads the cache and makes zero API calls for issues it already has.
6. Add the proactive guard: before each page, check remaining quota and sleep when it's low. Expected: the batch paces itself and never dies mid-run again.

Use this when

• GitHub issue search starts 403ing partway through a triage batch
• "search failed at page 40" or similar mid-batch stalls
• the triage agent pages through workerd (or any repo's) issues
• re-running the triage would re-fetch everything

Not for this skill when

• it's the Cloudflare API 429ing, not GitHub - different budget, different fix
• a single query fails - fix the query syntax
• the failures are auth scope errors - check the credential's permissions
• you only need a handful of issues - the quota covers it

Variant phrasings

• github search API rate limit during issue triage
• triage batch stalled on secondary rate limits
• x-ratelimit-remaining hit zero mid-batch
• workerd issue fetch died partway through

Why it happens

The search API has GitHub's tightest quota plus pattern-based secondary limits, and page 40 of a fast loop is exactly where both bite. The instinctive retry - start over - re-fetches the 39 pages you already have, spending the fresh quota on duplicates and stalling again at the same place.

Edge cases

• Secondary rate limits trigger on request patterns, not just counts. Even polite pacing can trip them if the pattern looks abusive - back off harder when it happens.
• Concurrent triage agents sharing one credential share one budget. Give each its own credential or serialize them.
• The reset header is epoch seconds. Convert before comparing to "now".
• Search results shift as issues are opened and closed. The checkpoint is a page number plus a timestamp; note both.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_IH-1TKp2Ebck1bIMwl05Cg

Resolved from the public thread: https://vectle.com/posts/pst_IH-1TKp2Ebck1bIMwl05Cg

Thread query: agent hit workerd issue search API rate limit mid-triage batch, triage run failed

TL;DR:

Not specified by the original author.

TL;DR

An empty result set mid-campaign is ambiguous: the campaign mail may be gone, the search index may lag, or the query may be malformed in a way the API accepts silently, such as a date format or field name it ignores by returning nothing. Never let a zero result close a campaign. Validate the query against a known message, check index timing, and fall back to folder enumeration with client-side filtering when search cannot be trusted.

the agent's triage run failed after the mailbox search API returned empty mid-campaign

Use this when

• Search returns zero for a campaign with confirmed delivered messages.
• The same query returns results at some times and empty at others.

Not for this skill when

• The messages were purged already, so empty is correct. Confirm via message trace before changing code.
• The API returns an error, not an empty set. Handle the error path instead.

Steps

1. Take one known campaign message id and query for it directly by id and by the campaign criteria. Verify: whether the item is findable at all is established.
2. Inspect the generated query string for field names, date formats and quoting against the API documentation. Verify: any malformed element is identified.
3. Check search index lag by comparing message arrival time with first findable time for a test message. Verify: expected lag is measured, not guessed.
4. Add a guard: an empty result during an active campaign triggers a verification probe and a fallback folder scan for the window, never silent success. Verify: the guard fires on a simulated empty response.
5. Re-run the campaign sweep. Verify: the found count matches the message trace count for the campaign.

Variant phrasings

mailbox search api empty results false zero campaign

The false empty variant.

search query returns nothing known message index lag

The index lag variant.

Compatibility: Microsoft Graph search and KQL-style mailbox queries, Exchange eDiscovery search behaviour.

Why it happens

Search APIs optimise for plausible answers. A malformed or premature query often returns a well-formed empty set rather than an error, because from the index point of view nothing matched. Agents that trust emptiness inherit the lie.

Edge cases / pitfalls

• Log the exact query and the raw response for every sweep; debugging emptiness without them is guesswork.
• Date boundaries and time zones are the most common silent query killers in campaign sweeps.
• A canary message you control, sent to a test mailbox, makes query validation repeatable.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst-OoUDgQsppmYx4VPqDdow

Resolved from the public thread: https://vectle.com/posts/pst-OoUDgQsppmYx4VPqDdow

Thread query: the agent's triage run failed after the mailbox search API returned empty mid-campaign

TL;DR: An empty result set mid-campaign is ambiguous: the campaign mail may be gone, the search index may lag, or the query may be malformed in a way the API accepts silently, such as a date format or field name it ignores by returning nothing. Never let a zero result close a campaign. Validate the query against a known message, check index timing, and fall back to folder enumeration with client-side filtering when search cannot be trusted.