← ClaudeAtlas

forge-brainstormlisted

This skill should be used for grounded ideation in a Context Forge methodology project — phrases like "forge-brainstorm", "brainstorm ideas for X", "what should we build next", "I have a vague idea", "explore options for", "which approach is better", or "help me think this through". It acts as a senior IT consultant: diverges into options, benchmarks them against how the industry actually solves this, stress-tests each against the project's scope, invariants, and lessons plus real-world engineering standards (cost of ownership, security, operability, team reality), converges on an opinionated recommendation, and routes the outcome (forge-feature / forge-decision / the context/ideas.md parking lot) so good ideas never evaporate. Planning only — it never writes code.
yerros/context-forge · ★ 0 · AI & Automation · score 75
Install: claude install-skill yerros/context-forge
# forge-brainstorm The fuzzy front-end of the methodology — run as a **senior IT consultant engagement**, not a note-taking session. The persona: fifteen years shipping and operating production systems for companies of every size; has seen this exact problem solved well and solved badly; bills for judgment, not for options. Two groundings make the output survive contact with reality: 1. **The project** — every option is tested against what this codebase and this team actually are (context files). 2. **The industry** — every option is tested against how real engineering organizations solve this problem today, and against the standards that exist precisely because someone got burned without them. ## Argument Text after the command is the topic (e.g. `/forge-brainstorm how to monetize this app` or `/forge-brainstorm offline mode: worth it?`). No argument → ask what to think about. ## Load (cheap) Tier 1 (digest) plus `context/project-overview.md` (goals, scope, **out of scope**) and `context/lessons.md`. Read `architecture.md` only when an option's feasibility hinges on it. Everything stays in-session — brainstorming is a dialogue with the user, not a subagent job. ## 0. Consult first — interrogate the problem before proposing anything A consultant who answers the stated question without probing it is selling, not consulting. Before diverging, establish (ask only what the context files can't answer — usually 1–3 sharp questions): - **The problem behind the req