| name | create-learning-course |
| description | Use only when explicitly invoked for creating or revising public-safe learning courses, lesson plans, curricula, course workspaces, syllabi, exercises, glossaries, and learner progress records from a learner mission and trusted sources. Use when Codex needs to turn a topic, audience, resource set, or existing notes into a structured course with lessons, retrieval practice, citations, and publication/security review. |
Create Learning Course
Purpose
Use this skill to create practical learning courses that are grounded in a
learner mission, built from trusted sources, and safe to share publicly by
default. The output can be a full course workspace, a smaller syllabus, or a
single lesson that fits an existing course.
Invoke this skill explicitly with $create-learning-course. It disables
implicit invocation because course creation can create or revise many local
files.
The course model is mission-led: curated resources, small lessons, reusable
assets, reference documents, and learning records. This skill is owned by this
repository and adds a stricter publication and security review so skills and
generated courses do not expose sensitive material.
Invocation Policy
Explicit invocation required. Use only by explicit invocation with
$create-learning-course. Do not implicitly invoke this skill for broad
course-creation requests, because it can create or revise many local files.
When To Use
- Creating a new course from a topic, target audience, desired outcome, source
list, or folder of existing public-safe notes.
- Turning a learner mission into a syllabus, lesson sequence, exercises,
glossary, reference sheets, and review checkpoints.
- Revising an existing course workspace while preserving its mission,
terminology, source list, lesson numbering, and learner records.
- Creating public-safe training material from private or internal notes by
generalizing examples and removing sensitive details.
- Adding retrieval practice, spacing, interleaving, and feedback loops to
lessons that are currently only explanatory.
When Not To Use
- Do not use for direct implementation, refactoring, debugging, code review,
commits, PRs, publishing, or hosting. Use the appropriate implementation,
review, Git, publishing, or hosting skill instead.
- Do not use for live tutoring when the user wants an interactive lesson in
chat rather than reusable course artifacts.
- Do not use for general technical due diligence without a course output. Use
research first when the user needs authoritative investigation before
curriculum design.
- Do not certify high-stakes training, compliance status, professional
licensing, medical/legal/financial advice, or safety-critical procedures.
Recommend qualified expert review for those courses.
- Do not copy internal documents, customer material, private URLs, raw logs, or
proprietary examples into public course artifacts.
- Do not create a broad encyclopedia. Courses should be built around observable
learner outcomes and practice, not exhaustive topic coverage.
Inputs
- Topic, learner audience, target skill level, desired outcome, timeline, and
constraints.
- Trusted resources such as public URLs, books, papers, official
documentation, source folders, or user-provided notes.
- Existing course workspace paths containing files such as
MISSION.md,
COURSE.md, SYLLABUS.md, RESOURCES.md, lessons/, reference/,
learning-records/, assets/, or PUBLICATION-REVIEW.md.
- Output path or format constraints such as HTML lessons, Markdown lessons,
workshop outline, self-study course, or instructor-led course.
Required Reads
- Read
references/course-design-workflow.md before creating or substantially
revising a course workspace.
- Read
references/publication-safety.md before writing final course artifacts,
converting private material into public-safe examples, or reporting that a
course is ready to share.
- Use
assets/course-workspace-template/ when creating a new course workspace
or when an existing workspace is missing core files.
- Use
evals/trigger-prompts.md when reviewing or tuning trigger behavior.
Writes
- May create or update local course workspace files requested by the user or
implied by the selected output path, including
MISSION.md, COURSE.md,
SYLLABUS.md, RESOURCES.md, GLOSSARY.md, PUBLICATION-REVIEW.md,
lessons/, reference/, learning-records/, and assets/.
- May create optional private planning notes only when the user explicitly asks
for them, and place them outside the publishable course root, such as
<course-folder>.private/NOTES.md.
- May update this skill's source materials under the Learning Loop when the
current task contract allows source edits and the learning is reusable,
evidence-backed, and public-safe.
- Must not write to external services, LMS systems, websites, email, Slack,
Confluence, Git remotes, or publishing platforms unless the user explicitly
invokes a separate workflow for that action.
Process
- Clarify the mission.
- If the learner goal, audience, success criteria, or constraints are vague,
ask the smallest useful question before writing course artifacts.
- If enough information exists, state assumptions and proceed.
- Establish the output boundary.
- Use the user-provided path when present.
- Otherwise create or update a local course folder in the current workspace
using a safe slug such as
course-<topic>.
- Do not overwrite existing course files without reading them first.
- Gather and rank sources.
- Prefer official documentation, primary sources, reputable textbooks,
peer-reviewed material, source code, standards, and expert-authored
public resources.
- Use private or internal material only as input context. Convert it into
public-safe, generalized course content unless the user explicitly says
the course is private.
- Build the course structure.
- Create or update
MISSION.md, COURSE.md, SYLLABUS.md,
RESOURCES.md, GLOSSARY.md when useful, lessons/, reference/,
learning-records/, assets/, and PUBLICATION-REVIEW.md.
- Do not create private notes by default. If the user explicitly asks for
non-secret private planning notes, keep them outside the publishable course
root.
- Design lessons.
- Make each lesson teach one tight outcome and produce one tangible learner
win.
- Include a short explanation, cited source links, retrieval practice,
worked examples, exercises, feedback or answer keys, and next review cues.
- Reuse assets before creating new lesson-specific styles or widgets.
- Run the publication safety review.
- Check course files for secrets, private endpoints, customer data,
proprietary details, unsafe claims, missing citations, and license issues.
- Redact or generalize unsafe content before calling a course public-safe.
- Report the result.
- Summarize the course path, artifacts created or updated, sources used,
safety review outcome, validation run, and remaining review needs.
Idempotency
- Before creating a course workspace, check whether the target path exists and
read its core files instead of replacing them.
- Preserve existing lesson, reference, and learning-record numbering; add the
next sequential number for new files.
- Update existing course files in place when revising the same mission, and
create a learning record when the mission or course strategy changes.
- Reuse existing
assets/ components before adding new styles, scripts, or
widgets.
- If a previous publication review exists, update its status and notes rather
than starting a competing review file.
Failure Handling
- If the learner mission is too vague to produce useful course artifacts, ask
concise mission questions before writing files.
- If trusted public sources are missing, create a source gap in
RESOURCES.md
and avoid unsupported claims in lessons.
- If private context appears necessary, distinguish private course status from
write permission. Secrets, private endpoints, customer data, regulated data,
raw logs, and credential material remain never-write. Use generalized
examples unless the user explicitly approves non-secret private planning
notes outside the publishable course root.
- If a publication-safety scan finds a concrete risk, redact or generalize the
content before reporting the course as public-safe.
- If validation tooling is unavailable, perform manual static inspection and
report the skipped command.
Guardrails
- Generated skills and generated courses must remain public, generic, and
reusable unless the user explicitly marks a course private.
- Use placeholders such as
{API_TOKEN}, {PROJECT_ID}, {PRIVATE_ENDPOINT},
{CUSTOMER_NAME}, and {INTERNAL_SYSTEM} instead of real values.
- Never print, persist, or copy secrets, private endpoints, customer data,
internal hostnames, raw logs, credentials, certificates, or non-public URLs
into course files, skill files, examples, final responses, or task state.
- HTML lessons must avoid analytics, trackers, remote scripts, remote fonts,
and unreviewed third-party embeds. Prefer local CSS and plain HTML.
- Cite public sources for factual claims. Mark claims as unverified when a
trusted source does not support them.
- For high-stakes topics, create educational material only with explicit
limitations and a required expert-review item in
PUBLICATION-REVIEW.md.
Must Not
- Do not publish, upload, email, or otherwise distribute a course from this
skill.
- Do not overwrite existing course files without reading them first.
- Do not include secrets, private endpoints, customer data, internal hostnames,
raw logs, credential material, or non-public URLs in public course artifacts.
- Do not include secrets, private endpoints, customer data, regulated data, raw
logs, credential material, or non-public URLs in private notes either.
- Do not cite private material in public course files.
- Do not claim runtime trigger activation unless it was observed in a fresh
Codex surface.
Completion Criteria
- The course mission, audience, outcomes, constraints, and out-of-scope topics
are recorded or explicitly marked as assumptions.
- Course artifacts requested by the user exist and link to the relevant source,
lesson, reference, or review files.
- Lessons include practice, feedback, and review cues rather than explanation
only.
PUBLICATION-REVIEW.md reflects the current public-safe status and any
remaining expert, source, license, or security review needs.
- Validation or manual static inspection has been run and limitations are
reported.
Validation
- Validate the skill itself with the repo skill validators after edits.
- Validate generated course artifacts with a static publication review before
calling them public-safe.
- Use file-listing secret scans or local inspection patterns that do not paste
suspected secret values into the conversation.
- Treat browser rendering, live web publishing, external LMS uploads, and
third-party service calls as live validation. Run them only after explicit
user request and safety confirmation.
Learning Loop
When using this skill, capture durable, reusable, public-safe learnings back
into this skill's local source materials before completion when the current task
contract allows source edits. Update the narrowest appropriate surface:
SKILL.md for runtime rules, references/ for detailed guidance, assets/
for reusable templates, scripts/ for deterministic helpers, and README or
changelog entries for human-facing or release-note updates.
If the current task is explicitly read-only/report-only, or source writes are
outside this skill's task contract, do not edit skill sources; report why
source capture was skipped.
Do not capture secrets, private URLs, customer data, raw logs, one-off local
state, or unverified/vendor-specific claims. If a useful learning is not safe,
not evidence-backed, or outside this skill's scope, report that it was skipped.
Output Contract
Return:
- Course path and artifacts created or updated.
- Learner mission, audience, outcomes, and constraints used.
- Source basis and any unverified claims.
- Lesson sequence and practice model.
- Public-safety/security review result.
- Validation run and skipped live checks.
- Remaining expert, learner, or publication review needed.
References
references/course-design-workflow.md: course workspace design, lesson
sequence, learning records, and source inspiration/attribution notes.
references/publication-safety.md: public-safe course creation rules,
redaction guidance, secret handling, and final review checklist.
assets/course-workspace-template/: starter workspace files for new courses.
evals/trigger-prompts.md: should-trigger and should-not-trigger examples.