Repository navigation
Enforce a pre-spawn model budget for planning invocations #163
Description
Split from #6.
Problem
docs/plan-format.md requires "pre-spawn byte/model-token budgets". Planning has byte limits only: MAX_PROMPT_BYTES in core/planning-author.ts and the issue-text limits in github/issues.ts. createClaudeCommand in agents/policy.ts passes no turn or budget flag, so one planning invocation is bounded by wall-clock time and output bytes only, not by model tokens or turns.
Acceptance
- Choose the budget (for example
--max-turns, or a token estimate checked before spawn) and record why in the docs. - Enforce it before the agent process starts. Refuse a request over budget with a clear error.
- Add a test at the exact budget and one unit over it, including the argv that
createClaudeCommandbuilds.
Activity
Word the model-budget docs to their evidence (#163)
Tie the planning input budget to the author's limits; note text-based…
Check a prepared draft at the prompt limit against the model budget (#…
Split the over-budget size into prompt and schema (#163)
Name every part of the request in the over-budget advice (#163)
Say which part of the backstop claim a test covers (#163)
Match the turn-limit text anywhere; note the errors text reaches ever…
Say that other output with the turn-limit text gets the same advice (#…
Re-record planning fixtures with the --max-turns budget (#163)
Enforce a pre-spawn model budget for planning invocations (#163) (#196)
Admit nothing after close with no agent chosen either (#164)
Check the planning agent at startup and before planning requests (#164)…