sangrokjung/claude-forge

BigCode Evaluation Harness evaluates code generation models on HumanEval, MBPP, MultiPL-E, and other benchmarks with pass@k metrics. Use it to compare coding ability, multi-language support, or overall code generation quality.

Blind Spot Pass surfaces the unknown unknowns in a domain before you start work, so you can prompt and decide with enough context. Use it when you are new to a field or handing off unfamiliar work and need a brief briefing first.

Build System detects a project's build tool and runs the right build or test command. Use it when setting up, building, or testing projects with npm, Python, Gradle, Maven, Cargo, Go, or Make.

Cache Components gives guidance for Next.js Cache Components and Partial Prerendering. Use it when `cacheComponents: true` is enabled to decide what to cache, stream, tag, and invalidate in Server Components.

Claude Code Agent guides Claude Code projects with spec-first planning, context engineering, sub-agents, and post-dev verification. Use it when writing CLAUDE.md or spec.md, dispatching parallel agent work, or running the handoff and docs sync workflow.

Continuous Learning V2 turns Claude Code sessions into atomic instincts using hooks, confidence scoring, and background pattern detection. Use it to capture repeated behaviors and evolve them into skills, commands, or agents.

Debugging Strategies helps you reproduce bugs, narrow causes, and verify fixes with systematic techniques and debugging tools. Use it for crashes, performance issues, memory leaks, or unfamiliar codepaths.

Dependency Upgrade helps manage major package version changes with compatibility checks, staged rollout plans, and testing. Use it when upgrading frameworks, handling breaking changes, or automating dependency updates.

Eval Harness defines capability, regression, and human-review evals for Claude Code sessions. Use it to plan work before coding, run checks during implementation, and report pass@k results.

Extract Error Codes runs `yarn extract-errors` to find new React errors that need codes assigned. Use it when adding error messages or when warnings say an error code is unknown.

Frontend Code Review reviews `.tsx`, `.ts`, and `.js` files for checklist violations. Use it for pending changes or named files before code is submitted.

Humanize Korean cleans up translation-ese and AI-writing tics in Korean prose while keeping facts, numbers, proper nouns, and quotes unchanged. Use it for a single-pass rewrite of existing Korean text that needs to read more naturally.

Korean Character Count measures Korean text with deterministic grapheme, line, and byte rules for self-intros and form limits. Use it when you need exact counts that match the skill's contracts.

Korean Spell Check uses Nara/PNU spell-check surfaces to proofread Korean text and suggest focused corrections. Use it for long or sensitive drafts that need chunking and conservative edit suggestions.

LLM Evaluation Harness evaluates language models on 60+ academic benchmarks such as MMLU, GSM8K, and HumanEval. Use it to compare models, report results, or track training checkpoints over time.

Loop Forge turns a one-line recurring task into a reusable slash command. It classifies the loop, asks for missing details, adds an independent verifier and a hardstop, then previews the command in the user’s language.

Manage Skills analyzes session changes to find verify-skill drift, missing coverage, and stale references. It is used to update or create verify skills and keep project CLAUDE.md registrations aligned.

Prompts And Skills Lookup searches, installs, and improves AI skills and prompts through prompts.chat or the skills.sh CLI. Use it when looking up a skill, installing one, or refining a prompt.

Relay compresses the current session into a one-line /compact command and a baton prompt for the next task. Use it when handing off work, continuing after compaction, or starting a new session.

Review Loop runs an implement-then-check cycle with a fresh independent checker before a change is marked done. Use it for behavioral changes, especially when you need an adversarial second pass on the current revision.

Security & Compliance guides security planning, threat modeling, incident response, and audit preparation for frameworks like SOC2, ISO27001, GDPR, and HIPAA. Use it to design, harden, monitor, and improve secure systems throughout the SDLC.

Security Pipeline runs CWE Top 25 checks, STRIDE analysis, and pre-commit vulnerability scanning when sensitive code changes or security commands are used. It helps gate commits and review fixes for common security issues.

Session Wrap runs at the end of a coding session to collect changes, detect documentation updates, repeated patterns, learning points, and follow-up tasks. It merges subagent results, lets the user choose what to act on, and records the outcomes.

Skill Factory analyzes session work, checks for duplicate patterns, and turns reusable workflows into new Claude Code skills. Use it when you want to extract a repeated task, make it a skill, or automate skill creation.

Strategic Compact suggests manual /compact at logical workflow boundaries instead of waiting for arbitrary auto-compaction. Use it to preserve context between planning, implementation, and debugging phases.