Learning Partner¶

The learning partner is woven into learnship, not bolted on. It fires at every phase transition when learning_mode: "auto" (the default), offering 2–3 contextually matched actions based on what just happened. You can also invoke any action at any time.

Core principle: Fluent answers from an AI are not the same as learning. Every action here makes you do the cognitive work: with support, not shortcuts.

How it activates¶

{
  "learning_mode": "auto"    // offered automatically at phase transitions (default)
  "learning_mode": "manual"  // only when you explicitly invoke @agentic-learning
}

On Windsurf, skills are natively invoked with @agentic-learning [action].
On all other platforms, reference the skill explicitly: use the agentic-learning skill: [action].

All 11 actions¶

learn: Active retrieval¶

@agentic-learning learn [topic]

You explain the topic first. The agent listens, then fills gaps with targeted follow-ups. This is deliberate retrieval practice: the most evidence-backed learning technique in cognitive science.

Best time to use: After research-phase, after debugging, when a new domain concept was introduced.

quiz: Active recall testing¶

@agentic-learning quiz [topic]

3–5 questions, one at a time, with formative feedback after each answer. Questions progress from recall to application to synthesis.

Best time to use: After execute-phase while implementation is fresh, after research-phase to test domain retention.

reflect: Structured reflection¶

@agentic-learning reflect

Three-question structured reflection: What did I learn? What was the intent? What gaps remain? Designed to be done in under 5 minutes: short enough to actually do it, rigorous enough to be useful.

Best time to use: After execute-phase completes.

space: Spaced review scheduling¶

@agentic-learning space

Identifies concepts from the current session and schedules them for spaced revisit. Writes a structured review plan to docs/revisit.md with suggested review dates based on forgetting curve intervals.

Best time to use: After verify-work passes, after pause-work (before ending a session).

brainstorm: Collaborative design dialogue¶

@agentic-learning brainstorm [topic]

A Socratic design conversation before committing to an approach. The agent asks probing questions, surfaces alternatives you haven't considered, and helps you find blind spots early.

Best time to use: After new-project, after discuss-milestone, before locking in a major architecture decision.

struggle: Productive struggle with hints¶

@agentic-learning struggle [topic]

You attempt to solve a problem from scratch. The agent provides a graduated hint ladder: giving only what you need to keep moving, never the full solution until you've genuinely tried.

Best time to use: After debug, after quick (when the task was tricky), when you want to cement a pattern you just used.

either-or: Decision journaling¶

@agentic-learning either-or

Records the decision paths considered, the choice made, the rationale, and expected consequences. Builds a searchable record of your reasoning that future phases (and future you) can reference.

Best time to use: After discuss-phase, after any significant architectural choice, after quick tasks with meaningful design decisions.

explain-first: Oracy exercise¶

@agentic-learning explain-first [topic]

You explain the concept or approach in your own words before seeing any reference material. The agent gives structured feedback on accuracy, completeness, and gaps.

Best time to use: After plan-phase (before executing), after research-phase, any time you want to test understanding before acting on it.

explain: Comprehension log¶

@agentic-learning explain [topic]

A deeper explanation exercise that writes to docs/project-knowledge.md: a persistent log of what you understand about how the project works. Good for onboarding, knowledge transfer, and future reference.

Best time to use: After significant phases, before handing off work, when building for future maintainability.

interleave: Mixed retrieval¶

@agentic-learning interleave

Active recall across multiple topics from different phases or sessions. Interleaving (mixing topics during review) is consistently shown to produce better long-term retention than blocked review.

Best time to use: After execute-phase when the phase covered multiple distinct concepts, at the end of a milestone.

cognitive-load: Scope decomposition¶

@agentic-learning cognitive-load [topic]

Breaks an overwhelming concept or task into working-memory-sized chunks. Uses chunking and progressive disclosure to make large scopes approachable.

Best time to use: After plan-phase when the scope feels overwhelming, before tackling a large or complex phase.

Which action when¶

Platform availability¶

The science¶

The actions in @agentic-learning are grounded in established cognitive science:

• Retrieval practice (learn, quiz, explain-first): actively recalling information produces stronger long-term memory than re-reading
• Spaced repetition (space): reviewing at increasing intervals exploits the forgetting curve for efficient retention
• Interleaving (interleave): mixing topics during practice produces better transfer and discrimination than blocked study
• Generation effect (struggle): generating answers (even wrong ones) before seeing the correct answer produces stronger encoding
• Elaborative interrogation (brainstorm, either-or): explaining why and how strengthens schema formation

Based on @FavioVazquez/agentic-learn.