Set or update a sub-project's curriculum

This skill builds the teaching plan for a learning sub-project. The curriculum is a reference document the agent consults when proposing what to study next — it does not force the order, but it makes the route through the material explicit and reviewable.

When to use

The user has bootstrapped a sub-project (via the bootstrap skill) and dropped source materials into .studyenv/<name>/source-materials/. They want to organize that material into an ordered teaching plan. Phrases: "set curriculum for tasep", "update the curriculum for spanish-b2", "plan how we should work through these papers".

This skill is not a session — it does not update PROGRESS.md. It only writes the curriculum file.

What you need from the user

Required:

• The sub-project name to act on. If the user did not name one, ask — never default to "all sub-projects" or guess.

If multiple sub-projects exist in the harness, prefer asking over assuming.

What to read

Before writing or updating the curriculum, load context:

1. The sub-project's .studyenv/<name>/AGENTS.md, falling back to .studyenv/<name>/CLAUDE.md for older projects — for Domain:, Language:, learning goals, and Tools & Materials.
2. The harness conventions at ../../reference/conventions.md — for the curriculum format and status legends.
3. The sub-project's .studyenv/<name>/ai-agent-materials/ — especially any existing curriculum.md and source extractions (concept maps, copied excerpts).
4. The sub-project's .studyenv/<name>/source-materials/ — the canonical source content.

If .studyenv/<name>/ai-agent-materials/ does not exist, create it. Concept maps and excerpts are added here over time; this skill primarily writes curriculum.md.

What to write

Render or update .studyenv/<name>/ai-agent-materials/curriculum.md with the structure documented in ../../reference/conventions.md (Curriculum format section):

• An overview paragraph — what this sub-project teaches and the intended end state.
• A topics section: an ordered list of topic entries, each with
  • a stable id (T1, T2.a, …) so PROGRESS.md can cross-reference
  • a one-line description
  • prerequisites (other topic ids; none if root)
  • suggested session type — one of the four core types theory, practice, role-play, onboarding (or both for theory + practice). theory/practice/role-play are run via start-session; onboarding via the dedicated onboard-session skill. Where the active overlay flavors role-play, the curriculum may name the flavor — simulation (Domain: speech-therapy), defense (Domain: academic-research), review / interview (Domain: coding). See ../../reference/conventions.md → Session types vs. flavors.
  • source references — page numbers, section anchors, URLs
  • exercise hooks — short outlines of proposed practice/theory exercises; the actual exercise artifacts live in work/, created at session time per the active domain overlay.

If curriculum.md already exists, update rather than overwrite:

• Integrate any newly added source materials.
• Fold in user-provided corrections or additions.
• Re-check dependencies and ordering.

The curriculum is reference-only. It does not constrain start-session or onboard-session — the agent may consult it when proposing a route, but is free to deviate.

Stay source-faithful; label any departures

This skill produces a source-faithful curriculum by default — every topic should be traceable to the user's source-materials/. If the source assumes a prerequisite it doesn't itself cover, you have two options: (a) name the prereq as a gap to fill at session time and don't add a curriculum topic for it, or (b) add a short bridging topic and label it [ext] per the External-source labeling rules in ../../reference/conventions.md. Do not silently weave training knowledge into a curriculum that's supposed to track the user's source. If the user wants a curriculum that genuinely shifts level (simpler / more advanced), they should use adjust-level instead, which is explicitly built for that.

Source-material handling

When extracting material from sources, build intermediate fixed representations under .studyenv/<name>/ai-agent-materials/ alongside curriculum.md:

• Terse concept maps — with references back to the source (page numbers for PDFs, section/heading anchors or URLs for other formats).
• Copied verbatim excerpts — sections, examples, definitions for explicit use during sessions. Keep these in the source's original language regardless of the sub-project's Language: setting.

Do not duplicate exercise content here; the curriculum's exercise hooks point at work/, but work/ is populated at session time by start-session.

Why these rules

• The curriculum is the explicit teaching plan, separate from PROGRESS.md (which tracks what's been done). Splitting plan from progress lets the user revise the plan without rewriting history.
• Topic ids are stable identifiers so PROGRESS.md can record T1: exercised without restating the topic description — and so reordering the curriculum doesn't break those references.
• Updating rather than overwriting matters when the user has been corresponding about plan revisions; the existing curriculum may carry their edits.

Related skills

• bootstrap mints the sub-project.
• start-session, onboard-session consult the curriculum (among other inputs) when proposing what to study next.