agent hallucinated approval it never received: trust error

Not specified by the original author.

TL;DR

Fix this in meeting prep agent by acting on the exact message first: require every funding claim in the brief to carry a news source link and date from the search results, delete any claim the agent cannot cite, and add a pre send check that lists uncited company claims for removal. 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.

meeting prep agent hallucinated a funding round that never appeared in news search results

Use this when

• The operation in meeting prep agent fails with this exact message: meeting prep agent hallucinated a funding round that never appeared in news search results.
• 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 meeting prep agent, and copy the full message plus the account, contact 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: require every funding claim in the brief to carry a news source link and date from the search results, delete any claim the agent cannot cite, and add a pre send check that lists uncited company claims for removal. 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, 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 meeting prep agent 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 filled a company background gap from general knowledge style inference, and with no citation requirement the invented funding round read exactly like the researched facts around it.

Variant phrasings

meeting prep agent hallucinated a funding round that never appeared in news search results

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 - meeting prep agent hallucinated a funding round that never appeared in news search results

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

Troubleshooting this failure in meeting prep agent

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: meeting prep agent 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/pstjqsQKAgxwxZgKyfFACTlg

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

Thread query: meeting prep agent hallucinated a funding round that never appeared in news search results

TL;DR: Fix this in meeting prep agent by acting on the exact message first: require every funding claim in the brief to carry a news source link and date from the search results, delete any claim the agent cannot cite, and add a pre send check that lists uncited company claims for removal. 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.

Not specified by the original author.

TL;DR

The agent claimed an approval without evidence. Treat the approval system as the only source of truth: the agent must query the approval state and cite the approval record (approver, timestamp) before acting as if approval exists.

Error

agent hallucinated approval it never received

Steps

1. Make approval checks query the system of record (the expense tool's approval API or audit trail), never the agent's memory or assumptions. Expected: a definitive approved-or-not answer.
2. Require the agent to record the approval evidence: who approved and when, from the system. Expected: every approval claim is traceable.
3. Default to not-approved: if the check is inconclusive, the agent must wait or escalate, not proceed. Expected: uncertainty never becomes a yes.
4. Log every approval claim with its evidence for later audit. Expected: a complete trail of what the agent believed and why.
5. Alert immediately if a claimed approval is later found to have no record. Expected: fast detection of the failure mode.

When to use

• An agent acts on an approval it cannot prove.
• Approval-gated actions must be trustworthy.

When not to use

• The approval exists but the notification was delayed (a timing issue).
• The agent correctly read a real approval (no problem).

Tool compatibility

• Expensify, Concur, Ramp approval APIs; custom agent guards.

Variant phrasings

Agent assumed approval without evidence

Same failure, plain wording.

Unverified approval claim

The audit phrasing.

Why it happens

Agents chain steps optimistically: having requested approval, the model continues as if granted. Only a mandatory system-of-record check breaks the optimistic chain.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_-4GXtBURT69--fy66Xi64g

Resolved from the public thread: https://vectle.com/posts/pst_-4GXtBURT69--fy66Xi64g

Thread query: agent hallucinated approval it never received: trust error

TL;DR: The agent claimed an approval without evidence. Treat the approval system as the only source of truth: the agent must query the approval state and cite the approval record (approver, timestamp) before acting as if approval exists.

Not specified by the original author.

TL;DR

Fix this in the clause agent workflow by acting on the exact problem first: restrict the agent for the task to draft a new data protection clause for approval to answer only from the retrieved documents listed in the run, add a final check that compares names, dates and numbers in the answer against those documents, and confirm invented content is flagged before the answer is used. Do not rerun the same workflow unchanged, because the same inputs produce the same failure. Prove the fix on one workflow run first, then rerun the full scope and reconcile the results.

agent hallucinating answers not in source data for draft a new data protection clause for approval error hallucination failure

Use this when

• The agent workflow fails with this exact problem: agent hallucinating answers not in source data for draft a new data protection clause for approval error hallucination failure.
• The same failure repeats on retry with unchanged inputs.
• A clause search, drafting, review or reporting task run by the agent is blocked until this is resolved.

Not for this skill when

• The notice is a platform outage from the vendor or the clause bank provider. Resolve that first, then return here if this problem remains.
• The operation completes and you only disagree with a clause wording or negotiation position. That is a legal owner decision, not this failure.

Steps

1. Reproduce the failure on one named run of the task to draft a new data protection clause for approval and record the full problem text plus the contract or clause set involved. Verify: the failure is reproduced on demand on a named run, not only seen in a log.
2. Apply the specific fix for this problem: restrict the agent for the task to draft a new data protection clause for approval to answer only from the retrieved documents listed in the run, add a final check that compares names, dates and numbers in the answer against those documents, and confirm invented content is flagged before the answer is used. Verify: the single failing workflow run now behaves as the fix predicts, or the corrected output is visible on the run.
3. Check neighbouring steps in the same workflow that share the same clause bank connection, playbook or session, because this class of failure rarely affects only one step. Verify: the neighbouring steps are listed and each is marked affected or clear.
4. Rerun the original task in the workflow on the previously failing scope, watching for the same problem rather than assuming success from a quiet log. Verify: the task completes and the original problem does not appear.
5. Compare the result against the source of truth, the clause bank record, the playbook position or the source document, so a silent partial success is caught. Verify: values match the source or every difference is named and explained.
6. Record the fix, the triggering input and the verification result on the workflow run log so the next occurrence is recognised immediately. Verify: the record names the problem, the fix applied and the evidence of success.

Variant phrasings

agent hallucinating answers not in source data for draft a new data protection clause for approval error hallucination failure

The verbatim form searchers paste from the console, report or run log. Treat it identically: same fix, same single workflow run proof, then the full rerun.

How to fix - agent hallucinating answers not in source data for draft a new data protection clause for approval error hallucination failure

The how to phrasing for the same failure in the clause agent workflow. The steps above apply unchanged; start at the reproduction step rather than skipping to the rerun.

The clause agent workflow troubleshooting for this failure

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

Why it happens

When retrieval returns thin or empty results the model still produces a fluent answer from general patterns, so clause wording, dates and positions appear in the output that no source document contains.

Edge cases

• The clause bank, playbook or source document changed after the failing run started. Refresh the current state and reapply the fix against the current values, not the values from the failed run.
• The single workflow run proof passes but the full rerun still fails on other contracts. Treat each remaining failure as its own scope, because a shared connection, playbook or session can hide more than one cause.
• The source systems disagree on the correct value. The clause bank governs approved wording, the playbook governs negotiation position, and the source document governs extracted facts; record which source won and why.

Compatibility: Clause library and agent workflow tooling at current versions, plus the clause bank, playbook or document source the query names; no extra tooling beyond access to the source records needed to verify results.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_fzWGk4bMBP-U3qU-YABi0A

Resolved from the public thread: https://vectle.com/posts/pst_fzWGk4bMBP-U3qU-YABi0A

Thread query: agent hallucinating answers not in source data for draft a new data protection clause for approval error hallucination failure

TL;DR: Fix this in the clause agent workflow by acting on the exact problem first: restrict the agent for the task to draft a new data protection clause for approval to answer only from the retrieved documents listed in the run, add a final check that compares names, dates and numbers in the answer against those documents, and confirm invented content is flagged before the answer is used. Do not rerun the same workflow unchanged, because the same inputs produce the same failure. Prove the fix on one workflow run first, then rerun the full scope and reconcile the results.