Distillation

Date: Sep 9, 2026

• 🔴 (14:09) Review-target verification observed committed HEAD bbaa7b3b722b87f1bccfe84063269cbbcabe2741 and git diff | sha256sum value 092bd817573c1e0ea81e71472a9da2f38251dd6736331152b8325526527c3479, matching the expected immutable target.
• 🟡 (14:09) Repository root /home/byk/Code/getsentry/cli-pr-1558 contained 18 entries: .craft.yml, .git, .github/, .gitignore, .lore.md, .nojekyll, .npmrc, AGENTS.md, apps/, CHANGELOG.md, codemods/, LICENSE.md, node_modules/, package.json, packages/, pnpm-lock.yaml, pnpm-workspace.yaml, and README.md.
• 🟡 (14:09) Applicable instruction files found at /home/byk/Code/getsentry/cli-pr-1558/AGENTS.md and /home/byk/Code/getsentry/cli-pr-1558/packages/cli/AGENTS.md.
• 🔴 [enforced-workflow] (14:09) User directive from /home/byk/Code/getsentry/cli-pr-1558/AGENTS.md: always load repo-setup before situation skills.
• 🔴 (14:09) Repository instructions identify the agent as “Jared (Outpost agent),” direct autonomous work to /workspace/repo, and define the pipeline as triage → explore → plan → implement → review → ship; worker is a deprecated alias of implement.
• 🔴 (14:09) Repository model assignments: primary Jared uses Claude Opus 4.8 for heavy triage/plan/review and xAI Grok 4.3 for light triage/plan/review; explore uses OpenAI gpt-5-mini; implement uses Moonshot kimi-k2.7-code; ship uses xAI Grok grok-build-0.1.
• 🔴 (14:09) Repository instructions state operator turns labeled New operator chat or Operator guidance: skip triage and should be treated directly as the task, with the answer returned in the conversation.
• 🔴 (14:09) Repository instructions state long-term project knowledge lives in .lore.md when present; target repositories require reading their AGENTS.md or CONTRIBUTING.md first.
• 🔴 (14:09) Repository instructions state skills live under .agents/skills/ and are generated from the canonical skills/ tree by scripts/sync-skills.mjs.