[caveman] Optimize instruction verbosity — adr-writer, contribution-checker (2026-07-03)#43266
Conversation
…hecker Trim prose that restates adjacent code/template signal without helping agents generate valid AW source. Co-Authored-By: Claude <noreply@anthropic.com>
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
|
Thanks for the caveman optimization run on One mechanical flag to note:
If you'd like to address this for future caveman runs, here's a prompt: Warning Firewall blocked 1 domainThe following domain was blocked by the firewall during workflow execution:
network:
allowed:
- defaults
- "patchdiff.githubusercontent.com"See Network Configuration for more information.
|
|
🎉 This pull request is included in a new release. Release: |
Summary
Reduces instruction verbosity in two agent definition files by consolidating redundant prose while preserving all actionable constraints.
Changes
.github/agents/adr-writer.agent.md.github/agents/contribution-checker.agent.mdWhere:list of 9 individual field-description bullets with aField values:list of 6 bullets.focused,deps,testsonto a single line.titleandauthorfield descriptions from the output spec.Impact
adr-writer.agent.mdcontribution-checker.agent.mdNo Go code, compiled artifacts, or tests changed. Pure instruction-text optimization to reduce token consumption in agent prompts.