From forge
Ideation and alternative exploration before architecture. Generates multiple approaches with tradeoffs, helps the user choose or combine, and produces a brainstorm artifact that /architect consumes. Use --grill to stress-test an existing plan instead of generating new approaches. Use before architecture when exploring options — triggered by 'brainstorm', 'explore alternatives', 'what are the options', 'ideation', 'compare approaches', 'grill my plan', 'stress test this', 'challenge this plan'.
How this skill is triggered — by the user, by Claude, or both
Slash command
/forge:brainstormThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Generate 3-5 genuinely distinct approaches (at least one unconventional/contrarian), surface honest tradeoffs, and help the user choose before committing to architecture. Output is a brainstorm artifact for /architect. Never pick for the user. Never write implementation code.
Generate 3-5 genuinely distinct approaches (at least one unconventional/contrarian), surface honest tradeoffs, and help the user choose before committing to architecture. Output is a brainstorm artifact for /architect. Never pick for the user. Never write implementation code.
Read $ARGUMENTS. If --grill flag present, set GRILL_MODE=true and strip it. If no arguments remain, ask what to explore.
Before exploring solutions, verify you're solving the right problem (skip obviously answered):
Wait for answers; reframe if they reveal a different problem. If GRILL_MODE=true, skip Steps 1-4 and proceed to Step 0.5.
Only runs with --grill. Stress-tests an existing plan instead of generating approaches.
Setup: Read codebase, tech stack, existing patterns, /memory-recall. Read the user's plan from $ARGUMENTS or ask. Identify key decision points to interrogate.
Interrogation Loop (one question at a time per decision point):
Categories (skip settled): Scope boundaries, Assumptions, Failure modes, Integration points, Missing pieces, Ordering risks.
Termination: Default cap 10 questions, then summarize and write artifact. "Keep going" extends 10 more; "enough"/"looks good" stops immediately. Stop early if out of genuine questions. Skip Steps 1-4, proceed to Step 5 with grill artifact format.
Gather context: codebase structure, patterns/constraints, related features, /memory-recall. Restate the core problem in 1-2 sentences and confirm with the user.
Generate 3-5 distinct approaches (not minor variations; at least one unconventional/contrarian). For each: Name, Description (2-3 sentences), Pros (2-3), Cons (2-3), Effort (L/M/H), Risk (L/M/H), Best for. Show summary table then full detail.
Wait for user reaction. Offer side-by-side table or hybrid possibilities if they want deeper comparison. Once they pick: sharpen the choice, note what was rejected and why (context for /architect), capture constraints discovered.
Write to .forge/brainstorm/[task-name-slugified].md (create dir if needed).
Grill fields: Title, Date, Mode (Grill), Plan summary, Decisions Confirmed, Risks Identified, Plan Changes, Open Questions, Constraints, Next: /architect. Normal fields: Title, Date, Problem, Approaches (Name/Desc/Pros/Cons/Effort/Risk each), Selected + Rationale, Rejected + Reasons, Constraints, Next: /architect.
Report selected approach, rationale, artifact path. Suggest /architect as next step.
Routing: See
skills/shared/workflow-routing.mdfor next-step guidance.
skills/shared/compliance-telemetry.md. Log violations via scripts/compliance-log.sh per shared protocol. Keys: insufficient-approaches, implementation-in-brainstorm, choice-not-presented (all major).skills/shared/rules.md.npx claudepluginhub bakirp/forge --plugin forgeGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.