| name | help |
| description | Explain the agentic-study-environment plugin — what it does, the lifecycle skills it ships, the domain overlays, the session types, and the key conventions — or a specific one of those if the user names it. Use this skill whenever the user asks for help with the study harness itself rather than asking to study something: phrases like "help with agentic-study-environment", "what can the study harness do", "how does this plugin work", "list the study skills", "how do I use set-curriculum", "what domains are there", "explain session types", or the /study-help command. With no topic, give the full overview; with a named skill, overlay, session type, or convention, explain just that. Do NOT use this to bootstrap, set a curriculum, or run a session — those are the lifecycle skills. |
agentic-study-environment help
Explain the agentic-study-environment harness to the user. This skill is the plugin's built-in guide — it answers "what is this and what can it do?" and "how do I use X?" so the user doesn't have to read the docs tree first.
This skill never creates or modifies any .studyenv/ files — it only explains. If the user actually wants to study, hand off to the lifecycle skill they need (bootstrap, set-curriculum, start-session, onboard-session, stop-session, adjust-level).
How to respond
Look at what the user asked for (the $ARGUMENTS from /study-help, or the topic named in natural language):
- No topic → give the full overview below, then point to the docs and the getting-started path.
- A specific topic (a skill, a domain overlay, a session type, or a convention like "tracking", "language", "status", "curriculum") → explain just that one: what it is, when it applies, and where it's defined. Read the matching
docs/<skill>.md, domains/<name>.md, or reference/conventions.md section and fold in the specifics. If ambiguous, list the close matches and ask.
Keep it skimmable: one-line answer first, then detail. Summarize and link rather than dumping file contents.
Full overview (use when no topic is named)
agentic-study-environment is a coding-agent learning harness: you bring a topic and source material (a paper, a book, a syllabus, a clinical case load, anything), and the agent becomes a tutor that brackets the work into focused sessions, keeps a per-project curriculum and progress log, and adapts its teaching shape to the domain you're studying via swappable overlays. It's a harness, not a course — there's no built-in content.
Everything it generates lives under a single .studyenv/ directory at the project root, organized as one or more sub-projects (one topic each). The host project's own files are never touched.
Lifecycle skills (the cross-agent surface)
| Skill | Trigger | What it does |
|---|
bootstrap | "bootstrap a project for X" | Mint a new sub-project (.studyenv/<name>/ with AGENTS.md, CLAUDE.md, PROGRESS.md, source-materials/); register it in .studyenv/PROGRESS.md. |
set-curriculum | "set curriculum for X" | Build or update .studyenv/<name>/ai-agent-materials/curriculum.md from the source materials, source-faithfully. |
start-session | "start session", "let's work on X" | Begin a bracketed session (theory / practice / role-play, with the active overlay's flavor); load the domain overlay. |
onboard-session | "onboard me on this codebase", "walk me through this repo" | Begin a bracketed onboarding session: survey an existing artifact, walk it part by part, reproduce a recent change. |
stop-session | "stop session", "wrap up" | Record progress to the sub-project + .studyenv/PROGRESS.md, and summarize what was covered. |
adjust-level | "simplify this", "make it harder", "build up to this paper" | Rewrite the curriculum at a different level, pulling in external material with strict source labels. |
help | "help with the harness" | This guide. |
Skills trigger from natural language — you rarely name them. On Codex they're exposed as slash commands (/agentic-study-environment:bootstrap, /agentic-study-environment:help, …); on Claude Code the /study-help command is a thin convenience over this skill.
The lifecycle loop
┌─ start-session ─┐
bootstrap → set-curriculum ─┤ ├─ stop-session
└─ onboard-session ┘
▲
adjust-level (re-level whenever the difficulty doesn't fit)
Session types
Four core types (a type changes the session loop): theory, practice, role-play, onboarding. theory/practice/role-play are proposed and run via start-session; onboarding — getting up to speed on an existing artifact you didn't author — has its own dedicated skill, onboard-session, since its shape (read-only access, a fixed survey/walkthrough/reproduce protocol) differs enough to warrant a direct entry point. Both close through stop-session. Each domain overlay adds flavors (same loop, different scaffolding + review focus) — e.g. speech-therapy simulation and academic-research defense are role-play flavors; coding adds review / interview flavors. The full type-vs-flavor rule lives in reference/conventions.md.
Domain overlays
The teaching shape (scaffolding form, review focus, /work/ layout, special flavors) comes from the sub-project's Domain::
coding — stub-file scaffolding, idiomacy review, language-appropriate /work/ layouts, the codebase flavor of onboarding (via onboard-session), and review / interview role-play flavors.
speech-therapy — therapist–patient simulation (a role-play flavor) over theory and practice.
legal-documents — prescribed-form drafting scaffolds, analysis frames (case briefs, redlines), legal-precision/citation review.
academic-research — the research lifecycle: critical-reading and literature-synthesis frames, manuscript/proposal scaffolds, peer-review and venue-fit exercises, and a defense role-play flavor.
A sub-project with no Domain: falls back to a neutral default (theory + practice over markdown notes under /work/). New overlays are the highest-leverage contribution — see domains/ADDING_AN_OVERLAY.md.
Key conventions (quick reference)
- Language — default English; override per-sub-project (or globally via
.studyenv/AGENTS.md) with Language: <BCP 47 tag>. Structural tokens (status legends, field names, skill names) stay English.
- Topic status —
introduced · exercised · reviewed (record the highest reached).
- Sub-project status —
created · ready · in progress · blocked · stopped · finished.
- Tracking scope —
Tracking: global (default) registers a sub-project in the cross-project .studyenv/PROGRESS.md; Tracking: local-only keeps just the sub-project's own PROGRESS.md.
Full conventions: reference/conventions.md.
Getting started
- Bootstrap — "bootstrap a project called
tasep to learn the totally asymmetric simple exclusion process" (optionally with Domain:, Language:, Tracking:, Tools & Materials).
- Drop source materials into
.studyenv/<name>/source-materials/.
- Set the curriculum — "set curriculum for
tasep".
- Start a session — "start session" proposes a topic + type (
theory/practice/role-play); override as you like. To get up to speed on an existing artifact instead, say "onboard me on X" (onboard-session).
- Stop the session — "stop session" records progress and summarizes.
Where to go next
- Per-skill guides:
docs/README.md and the per-skill pages in docs/.
- Domain overlays overview:
docs/domains.md.
- Plugin overview, transcripts, install, FAQ, limitations:
README.md.
Resolving a named topic
Map the user's word before explaining:
- "bootstrap", "new project", "set up a topic" →
bootstrap (docs/bootstrap.md).
- "curriculum", "plan", "syllabus" →
set-curriculum (docs/set-curriculum.md).
- "start", "session", "study", "let's work on" →
start-session (docs/start-session.md).
- "onboard", "onboarding", "walk me through", "get up to speed", "codebase I didn't write" →
onboard-session (docs/onboard-session.md).
- "stop", "wrap up", "record progress" →
stop-session (docs/stop-session.md).
- "level", "too hard", "too easy", "simplify", "harder" →
adjust-level (docs/adjust-level.md).
- "domain", "overlay", "coding / speech-therapy / legal / research" →
docs/domains.md and the matching domains/<name>.md.
- "session type", "theory / practice / role-play / onboarding", "flavor" →
reference/conventions.md (Session types vs. flavors).
- "tracking", "global vs local", "language", "status", "curriculum format", "layout" → the matching section of
reference/conventions.md.
Then open the matching file and explain from it.