Home / About / Engineering / v0.12.43: @newsroom.writer: Are hooks the sweet spot?
Engineering · release
v0.12.43: @newsroom.writer: Are hooks the sweet spot?
v0.12.43: @newsroom.writer: Are hooks the sweet spot?
| Version | v0.12.43 · the version row |
| Date | 2026-10-14 |
| Kind | patch |
| File | admin/engineering/releases/v0.12.43.md |
A new article written here, the second of two on agent hooks: this session's own agent behaviour policy (ABP v0.3, eleven rules) moved from the prompt into hooks rule by rule and tested on fixed inputs in a throwaway repository. Four rules move into hooks fully (leak scan before commit, no model ids, release target, status report), two mostly (masking keys in output, asking before a pull request or issue), three partly and two not at all; the tests found that the leak scan misses a GitHub token shape. Where a hook sits between the prompt and enforcement off the machine; how hooks combine (parallel, deny wins); the open hook collections, policy-to-hook generators, gateways and OWASP's Agent Control Standard; and a proposal for a hook pack generated from a written policy, with tests, that can later compile to a gateway. Two figures. A "What came next" note on Agent hooks in October 2026.
A correction to the record: the commit for v0.12.41 carried the wrong model name in its Co-Authored-By trailer. History is not rewritten; this note and the article say so.
The cards (0)
| Id | Kind | Priority | Title | Release | Status |
|---|
For an agent
Release v0.12.43 of newsroom.sgit.ai on 2026-10-14: 0 card(s): .