All authors
Claude Skills by johntran777
github.com/johntran7774 skills0 installs1 views
- Grill MeUser-invoked interviewing skill (`/grill-me <topic>`). The AI relentlessly interviews the user about a plan or feature — one question at a time, each carrying a recommended answer, drilling into purpose, edge cases, data shapes, failure modes, scope boundaries and success criteria — until the human and the AI hold the same design concept. Looks facts up itself rather than asking, defers questions no conversation can settle to a prototype, and ends by playing the full design back for explicit ...Votes: 0GitHub stars: 6
- Tran Forge HistoryUser-invoked documentation skill (`/tran-forge-history [KEY-123]`). Reconstructs a completed (or in-flight) Tran Forge cycle from the evidence it left behind — the base branch's commit ledger, the role branches, the Gherkin artifacts, surefire and PIT reports, the requirements file's decision entries, and the real Codex review prompts recovered from `~/.codex/sessions` rollout logs — and renders it as a single self-contained local HTML page with an interactive pipeline diagram: click any stag...Votes: 0GitHub stars: 6
- Tran ForgeUser-invoked TDD pipeline skill (`/tran-forge`). Use to drive one feature through the Tran Forge cycle — optional Jira intake → Grilling (Phase 0: refine a vague requirement via /grill-me — one question at a time — or /grill-me-batch — all questions together — per the config's grill_mode; interactive, skipped when already well-formed) → Specifier (Gherkin, user-approved) → Coder (strict TDD, tidy-as-you-go, coverage to the bar) → three Codex reviews (design/architecture, then security/OWASP, ...Votes: 0GitHub stars: 6
- Grill Me BatchUser-invoked interviewing skill (`/grill-me-batch <topic>`). The batch form of /grill-me: the AI does every lookup first, then asks ALL of its questions together as one numbered list — grouped by rung (purpose, users, happy path, edge cases, data, failure modes, non-goals, success criteria), each with a recommended answer — and the human replies only with the numbers to change plus `rest: ok`. At most two follow-up rounds handle the follow-ups and contradictions that the answers caused, then ...Votes: 0GitHub stars: 6