require-max-output-tokens-in-call-loops

━━ when it runs

• PreToolUsebefore the tool call runs. a denial here means the command never executes at all.

watchesevery tool the event fires for — not narrowed to particular tools

this pack enforces: a match returns a real denial rather than being recorded and discarded. what the denial then does depends on the event above.

━━ install

this one is opt-in. taking the pack without flags leaves it off until you tick it, so install it by name if you want only this.

• just this policy$failproofai policies add chhhee10/cost-guard --policy require-max-output-tokens-in-call-loops⧉ copy
• the whole pack$failproofai policies add chhhee10/cost-guard⧉ copy
• turn just this off$failproofai policies remove require-max-output-tokens-in-call-loops⧉ copy
• pick from a list$failproofai policies add⧉ copy
• uninstall the pack$failproofai policies remove chhhee10/cost-guard⧉ copy

ships in chhhee10/cost-guard · v4e133099d12a · source ↗

━━ also in inference

• cap-frontier-eval-row-countAsk for a row cap before an eval loop runs a frontier model over a full dataset
• bound-inference-job-concurrencyRefuse unbounded fan-out on a job that makes one model call per item
• sample-before-embedding-or-finetuningAsk for a sampled run before a fine-tune or an embedding pass over a full corpus