forge-reviewlisted

# Forge Review Code review with fresh context. This skill is deliberately run without implementation context so it catches things the author missed due to familiarity bias. Works in three modes: - **Pipeline mode** — invoked with documentId, posts findings as issue comment, advances status - **Subagent mode** — spawned by forge-code during Step 10, returns findings to caller - **Standalone mode** — `/forge-review` with no args, reviews current branch diff ## Usage ``` /forge-review <documentId> # pipeline — review + post comment + set status /forge-review # standalone — review current branch diff ``` ## Review Process ### 1. Get the Diff ```bash git diff HEAD~N --stat git diff HEAD~N git log --oneline HEAD~N..HEAD ``` N = number of implementation commits (exclude previous review/fix commits). ### 2. Load Relevant Skills Detect tech stack from changed files, then load only what applies: - Strapi files (`content-types/`, `controllers/`, `services/`) → read `.claude/skills/strapi/SKILL.md` - Next.js files (`app/`, `components/`, `next/` imports) → read `.claude/skills/nextjs/SKILL.md` - Also read `forge/.forge/lessons.md` if it exists — past gotchas to check against ### 3. Review Checklist **Bugs & Logic** — wrong logic, null risks, race conditions, missing error handling **Security** — injection, credentials in code, missing auth, unsanitized input **Performance** — N+1 queries, unnecessary re-renders, memory leaks, unbounded data **TypeScript**