AI Engineering

Ijfw Recall surfaces relevant project memory at session start or when asked with recall-related prompts. It summarizes working and project memory entries and points to full text on request.

Learn automatically evaluates OmA sessions to extract reusable patterns like error fixes, workarounds, and conventions, then saves them as rules. Use it after sessions or when configuring the SessionEnd hook and learned rules.

Learning Coach guides one step at a time through real projects, code, tasks, or materials while adjusting to the learner’s current state. It is used when someone says to start, continue, or check progress on learning something concrete.

Ring Orchestrator enforces the Ring conversation protocol by checking for applicable skills, dispatching agents, and avoiding direct tool use. Use it at the start of a task or whenever a request calls for exploration, refactoring, or file work.

Risk-Adjusted Planning helps structure high-risk work with stage gates, pass/fail criteria, and rollback points. Use it for refactors, migrations, or other changes that need correctness and risk control before coding.

SRT Card Annotator adds contextual reference cards below subtitle lines in SRT files without changing timestamps or original text. Use it for educational, instructional, or Chinese subtitle workflows that need key points, warnings, tips, or list markers.

SRT Enhancer compares subtitle text with an origin.md reference to correct typos, standardize terms, and add spacing around English words and numbers. Use it to refine SRT files while keeping timestamps and structure unchanged.

Testing Skills With Subagents hardens discipline-enforcing skills by running RED-GREEN-REFACTOR scenarios on a subagent without the skill, then tightening the wording until the agent follows it under pressure. Use it when a skill could be rationalized away.

UltraGoal breaks complex, multi-step work into ordered micro-goals with durable progress tracking across sessions. Use it when a change spans many files or needs strict, goal-by-goal execution and verification.

Absolute Spec turns a fuzzy feature idea into a reviewed design spec, then stops. Use it to discuss, hand off, or review a change before implementation, with a short clarify pass and no build step.

Absolute Work guides AI coding agents through a gated SDLC from intake and design review to task decomposition, test-first execution, verification, and convergence. Use it for tickets, migrations, refactors, and other multi-step work that needs planning before code.

Agent Loop detects requests for iterative autonomous task execution and routes them to the right loop executor. Use it when a user wants to keep trying, fix until green, or run an explicit agent loop.