# BullyBearAI AFK implementation cycle Repository: jeanbrianc/BullyBearAI Epic: #1090 Scope: fresh-state audit and actionable AFK children within the epic Maximum runtime: 12 hours Maximum concurrent implementation workers: 2 Usage ceiling: the remaining configured account allowance; stop at its limit Additional infrastructure / paid-service spend: $0 for the implementation pass Progress destination: afk/run-ledger.csv plus issue and draft-PR updates Checkpoint cadence: after each slice and before any handoff or interruption These are editable example limits, not reconstructed settings from the overnight case. Choose an explicit usage cap in your account where available. If the agent cannot observe usage, it must report that limitation rather than promise automatic enforcement. Read the current repository rules, epic, issues, PRs, and delivery instructions. Refresh which tasks are AFK-ready, HITL-required, or blocked. Use separate worktrees for implementation workers and serialize overlapping edits. Do not create duplicate or out-of-scope work to fill the runtime budget. Implementation authorization: code, tests, docs, issue progress, branches, and draft PRs. No merges, AWS deployment, broker or credential access, notifications, approvals/preflight records, trading-flag changes, or orders. Deliverable: issue/PR/commit links, test commands and results, package artifacts, dependency-aware review order, failed or skipped checks, and exact HITL work still open. Next session: use afk/openai/review-handoff.md or afk/claude/review-handoff.md. Refresh GitHub and confirm the deployment target before applying that separate review/deployment authorization.