Forge-Agentic-Coding-CLI

Deep review — random exploration + cross-agent review, alternating. Finds integration bugs that self-review misses. Run every 30–60 min during active implementation.

Overview of Forge's Agentic Coding Flywheel methodology. Use to route between plan, bead, and code space. See FLYWHEEL.md at the repo root.

End-of-session ritual. Close finished beads, file new ones for remaining work, run verify, commit in logical groupings, push. Work is not complete until git push succeeds.

Start a new Flywheel planning round. Produce initial markdown plan, route to competing frontier models, synthesize hybrid, refine 4–5×. Use before any coding begins on a new feature.

Translate a polished markdown plan into self-contained beads in .beads/beads.jsonl. Every bead must be executable without reopening the plan. Use after the plan has converged.

Polishing pass over all open beads. Run 4–6 times. "Check your beads N times, implement once." Use after plan-to-beads and between refinement rounds.

Break the spell of local productivity. Ask whether completing all open beads actually closes the gap to the project goal. Use at milestones or when busy agents feel directionally off.

Remove telltale AI-writing patterns from README, docs, and user-facing text. Must be done line-by-line, manually. No regex. Use after agents write any user-visible prose.

Merge duplicate or excessively overlapping beads in .beads/beads.jsonl into canonical ones. Run after any batch where 20+ new beads were created.

Deep code review after beads have landed — random exploration + cross-agent review, alternating. Use every 30–60 min during active implementation, or after any epic lands.

Overview of Forge's Agentic Coding Flywheel methodology — plan space, bead space, code space. Invoke this skill when you need to know which Flywheel skill or prompt applies to the current phase.

Reset review context when bead polishing or code review has flatlined. Requires starting a brand-new Claude / Codex session. Use when you're seeing the same small edits round after round.

30 → 5 → 15 funnel for brainstorming new features on the existing Forge codebase. Use when adding capability to a mature install, not for greenfield.

End-of-session ritual — close finished beads, file new ones for remaining work, run verify, commit in logical groupings, push. Work is not complete until git push succeeds.

Start a new Flywheel planning round for a Forge feature. Writes the initial markdown plan to .flywheel/plans/ and gives instructions for sending it to competing frontier models. Use at the very beginning of a new feature or subsystem.

Synthesize competing plans from multiple frontier models into a single best-of-all-worlds hybrid. Use after /plan has produced 2+ competing plans in .flywheel/plans/.

Translate a polished markdown plan into self-contained beads in .beads/beads.jsonl. Treat this as a distinct translation problem, not task extraction. Use after /plan-synthesize produces a stable hybrid plan.

Polishing pass over all open beads. "Check your beads N times, implement once." Run 4–6 times. Use after /plan-to-beads and between refinement rounds.

High-level reality check when the swarm feels busy but directionally off. Use when lots of code is landing but the real goal still feels far away.

Remove AI-writing tells from README, docs, and user-facing text. Must be done line-by-line, no regex. Use after agents write any user-visible prose.

Scaffold a new Forge subagent (planner/architect/executor/reviewer/debugger/memory-style).

Scaffold a new model provider for Forge — class, registry wiring, providerEnum entry, local-catalog classification, router check, and a unit test with mocked transport.

Scaffold a new Forge tool end-to-end — implementation, registration, permission/risk metadata, sandbox routing, and a unit test.

Systematic debugging playbook for the Forge agentic loop (src/core/loop.ts and src/agents/executor.ts). Use when a task is stuck, looping, or emitting wrong tool calls.

Pre-release checklist for Forge — verify, version bump, changelog, docs, docker build, smoke test. Run before tagging v*.

Run Forge's canonical verification chain (format:check + lint + typecheck + build + test). Use before claiming any task complete or opening a PR.

Scaffold a new Forge subagent (planner/architect/executor/reviewer/debugger/memory-style). Use when adding anything to src/agents/.

Scaffold a new model provider for Forge — class implementation, registry wiring, providerEnum entry, local-catalog classification, and unit test with mocked transport. Use for anything under src/models/.

Scaffold a new Forge tool end-to-end — type, implementation, registration, permission/risk classification, and unit test. Use when adding anything to src/tools/.

Systematic debugging playbook for the Forge agentic loop (src/core/loop.ts and src/agents/executor.ts). Use when a task is stuck, looping, or producing wrong tool calls.

Pre-release checklist for Forge — runs verify, checks version bump, changelog, docs, and docker build. Use before tagging v*.

Run the full Forge verification suite (format:check + lint + typecheck + build + test). Use this before claiming any task complete or before opening a PR.

Scaffold a new Forge tool — implementation, registration, permission/risk metadata, sandbox routing, and a unit test. Use when adding anything to src/tools/.

Run Forge's canonical verification chain (format:check + lint + typecheck + build + test). Use before claiming any task done or opening a PR.