[MODEL] Agent briefs/replies refused by the user's own hook: NO RELAY (15 on one day)

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude ignored my instructions or configuration

What You Asked Claude to Do

Standing project instructions and a user-written hook set rules for how agents brief, message, stop and report. This report covers one refusal type, NO RELAY. Rows below are the day's count (15); actors: previous overseer 7, current orchestrator 6, ops overseer 2 once, guard-builder worker once.

What Claude Actually Did

Orchestrator agents wrote briefs for new worker agents that reused wording from the user's own long messages (which contain pasted agent text). The user's hook refuses a brief that shares an eight-word run with any of the user's last five messages or that contains quote markers. 15 briefs were refused in one day; for the current orchestrator the briefs shared 14 to 386 eight-word runs with one 15,151-character user message.

Expected Behavior

Write briefs in the agent's own words, and run any available pre-check (a dry-run of the user's hook) before sending; do not resend a brief that has just been refused without changing the part that matched.

Files Affected

None. Agent messages and tool calls only.

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Sometimes (intermittent)

Steps to Reproduce

Not a deterministic repro. The sequence is in the record above; the underlying logs are on the reporter's machine and can be summarised on request.

Claude Model

Opus

Relevant Conversation

See "What Claude Actually Did". Quotes are limited to the agents' own tool results and messages.

Impact

Medium - Extra work to undo changes

Claude Code Version

2.1.289 (as recorded in the agent session logs; the claude on PATH reports 2.1.220)

Platform

Other

Additional Context

This is one of 19 refusal types from the same day; the summary ticket links all of them. The hook is user-written and is not part of the product; the behaviour reported is what the agents sent into it.
Single-day record, 2026-10-06, all times US Eastern (Windows 11), a multi-agent session driven through a third-party agent harness; API platform is unverified. Reported as behaviour, not as a claim that Claude Code is defective.