Imported from Phonghigh/learning (AGENTS.md). Install upstream with npx skills add Phonghigh/learning. Copyright stays with the author.

Learning Archive Workspace

Purpose

This workspace turns raw self-learning notes into durable project knowledge and LinkedIn-ready content. The user may paste incomplete notes, chat transcripts, errors, design decisions, or code observations.

Default workflow

When the user says capture learning, log this, or pastes a learning without another explicit task:

1. Identify the project. Prefer an existing project under project/; create a project folder only when the project is clearly new.
2. Classify the dominant technical category using the existing archive categories. Keep project for project-journey material, not for every technical lesson.
3. Extract: problem, context, attempted approaches, evidence, chosen approach, trade-offs, result, and reusable lesson.
4. Write the durable learning note under project/<project>/learning/.
5. Update the relevant project index and the root INDEX.md only when the user explicitly asks to persist the note.
6. Offer or create a mind map when the learning contains multiple decisions, phases, or dependencies.
7. Offer or create a LinkedIn draft only when requested, or when the user explicitly asks for content conversion.

Classification rules

• project: project narrative, phase evolution, architecture journey, roadmap, milestone, or cross-cutting synthesis.
• backend-api: API contracts, service logic, backend data-shape decisions.
• database: schema, migration, constraint, query, persistence decisions.
• devops-infra: Docker, deployment, process lifecycle, environment, CI/CD, runtime operations.
• frontend-ui: UI state, rendering, component behavior, browser-side async flows.
• gis-mapping: maps, geospatial data, layers, projections, raster/vector behavior.
• Prefer one dominant category. Mention secondary topics in the note body instead of duplicating files.

Writing rules

• Preserve uncertainty: label inferred causes and unverified assumptions.
• Explain trade-offs and rejected approaches; do not write a changelog.
• Separate facts observed from interpretation.
• Use Vietnamese when the input is Vietnamese; keep technical identifiers in English.
• Never invent metrics, test results, URLs, or implementation details.
• LinkedIn drafts must teach one clear idea, use a concrete story, avoid generic motivation, and end with a practical takeaway.

File layout

project/<project>/
  INDEX.md
  project-overview.md
  learning/*.md
  mindmaps/*.md
  linkedin/*.md

Use kebab-case file names. Do not create Markdown outside E:\Learning\plans or E:\Learning\docs unless the file is part of this existing archive structure.

Available local skills

• .Codex/skills/capture-learning/SKILL.md
• .Codex/skills/mindmap-learning/SKILL.md
• .Codex/skills/linkedin-post/SKILL.md