agent-forge-harness

Run red-green-refactor in vertical slices, emphasizing behavior-first tests and harness quality gates.

Break a plan/PRD into independent vertical-slice Beads issues with explicit dependencies.

Synthesize current context into a PRD and file it as a Beads feature/epic issue.

Phase 2 of the Forge pipeline. Turn a research document into an implementation plan that is built around Beads issues and TDD. Reads plans/research/<slug>.md, designs the build, sets up red-green-refactor test strategy using the tdd skill, materializes a Beads epic/feature/tasks with dependencies and priorities, and defines demo/test checkpoints. Use when starting /forge-plan or the plan step of /forgemaster.

Fact-check an implementation plan against the real codebase and first-party documentation, flagging inaccuracies with what / why / evidence and a severity tier. Use when the user wants a plan reviewed or validated for accuracy, runs /review-plan, or wants a plan checked against real code before implementing. Accepts a Beads issue, a filesystem path, or a plan pasted into chat.

Phase 3 of the Forge pipeline. Execute the plan's Beads tasks with red-green-refactor TDD, committing per checkpoint and pausing at each natural stopping point so the user can run a live demo or test and see the work themselves. Use when starting /forge-implement or the implement step of /forgemaster.

Phase 1 of the Forge pipeline. Gather real context from the codebase, answer everything the code can answer through exploration, then grill the user one question at a time (with a recommended answer) for only what code cannot answer, and write a research document. Use when starting /forge-research or the research step of /forgemaster, or when the user wants a research-first investigation before planning a feature.

Meta-skill for creating Agent Forge skills under .claude/skills/ with SKILL.md, optional references/ and scripts/, and Bun scaffolds. Use when the user wants to add or author a skill, scaffold a new skill folder, or align skill docs with harness conventions (JSON script output, Beads for tasks).

Phase 4 of the Forge pipeline. Summarize the whole workload — what was built, which Beads issues closed, a before/after, and a step-by-step walkthrough the user can run to verify it themselves — then run quality gates, push, and open the PR. Use when starting /forge-ship or the ship step of /forgemaster.

Ultra-compressed communication mode that preserves technical accuracy while dropping filler. Use when the user explicitly says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman.

Stress-test a plan against existing domain language, code behavior, and durable decisions. Use when the user wants DDD/domain-model grilling, glossary refinement, terminology challenges, context documentation, ADR candidates, or plan validation against a repo's language.

Generate multiple radically different interface designs before implementation. Use when designing an API, module boundary, command surface, data contract, or when the user says "design it twice".

Set up Claude Code git safety guardrails that block dangerous git commands before execution. Use when the user asks to prevent git push, hard reset, git clean, destructive checkout/restore, or branch deletion from Claude Code hooks.

Triage GitHub issues through gh and a label-based state machine. Use when the user intentionally wants GitHub issue triage, incoming bug or feature review, issue workflow labels, needs-info follow-up, or ready-for-agent briefs.

Edit and improve article drafts by confirming a heading-based section plan, dependency order, then rewriting section by section with short paragraphs. Use when the user wants to revise or improve an article draft.

Surface architectural friction and propose deepening opportunities that improve testability and AI-navigability. Use when the user wants to improve architecture, find refactoring opportunities, consolidate tightly-coupled modules, or make a codebase easier for agents to navigate.

Migrate test files from `as` assertions to @total-typescript/shoehorn helpers. Use when tests need partial fixtures, intentionally invalid data, or safer replacements for TypeScript assertions.

Relentlessly stress-test a plan or design through a one-question-at-a-time interview that researches available context first and includes a recommended answer with every question. Use when the user asks to "grill me", wants a plan/design challenged, or needs decisions resolved before implementation.

Run a conversational QA session where the user reports bugs or issues, and file durable Beads issues using project domain language. Use when the user wants to report bugs, do QA, file issues conversationally, or mentions a QA session.

Create a Beads-first refactor plan through detailed interview, repo verification, alternatives, scope boundaries, testing decisions, and tiny working increments. Use when the user wants to plan a refactor, write a refactoring RFC, or turn a risky change into safe incremental work.

Search, create, and manage notes in an explicit Obsidian vault with title-case filenames, wikilinks, index notes, backlinks, and safe vault-root confirmation.

Scaffold course exercise directories with numbered sections, exercise variants, required readmes, and repo-specific validation. Use when creating exercise stubs, adding a course section, or safely renaming exercises.

Set up repository-local Husky pre-commit hooks with lint-staged, Prettier, and optional typecheck/test gates. Use when a user asks to add pre-commit hooks, configure lint-staged, enforce staged formatting, or add commit-time checks.

Investigate a reported problem, diagnose root cause, and file a Beads issue with a behavior-level TDD fix plan.

Extract a DDD-style domain glossary from conversation context, choosing canonical terms and flagging ambiguity. Use when the user asks to define domain terms, build a glossary, clarify terminology, create a ubiquitous language, or mentions domain modeling or DDD.

Produce a higher-level architecture and context map for unfamiliar code. Use when you need to understand how a file, module, feature, or issue fits into the bigger picture, including relevant modules, callers, callees, dependencies, and next context.

Multi-repo git operations — init, reset, status, branch management, refresh (rebase), stacked rebase. JSON output for agent consumption.

Register a new sub-repo in the knowledge base.

Grade and steer UI work using Anthropic-style design dimensions—coherence, originality, craft, and usability—for generators and evaluators.

Choose Beads issue priority (P0–P4, numeric, or named) from urgency, impact, and risk when creating or triaging work.

Add focused Bun unit tests for mission-critical behavior and edge cases — not blanket coverage.

Generate structured weekly status reports covering initiatives, progress, blockers, risks, and KPIs.