IBM SkillsBuild's Cybersecurity and Data Analytics certificates carry ACE credit recommendations and four community-college-system deployments. What they do not carry is any published way to put that content, or that completion data, inside a Canvas course. This brief separates what is documented from what is gated behind a partner agreement — and proposes three tiers of integration a college can start on this semester, without waiting on IBM.
Detail and evidence follow. These are the short answers I'd bring to the discussion, and I'll defend each one below.
Run link-out now and noncredit CDCP as the scale play, in parallel — not in sequence. The ASCCC documents exactly this pattern: colleges launch a not-for-credit course to meet an immediate industry need while developing the credit and noncredit versions with industry partners. Waiting for the curriculum cycle to finish before offering anything costs a year of students, and waiting for IBM to ship an LMS connector costs an unknown number of years. California is already inside this program — the Bay Area Community College Consortium was one of the four launch systems in fall 2024. The statewide question is not whether to start, it is how to extend from one consortium to 116 colleges.
I checked the primary sources directly rather than relying on summaries. Each item below is marked by what the documentation supports — not by what seems likely.
ACE conducted a Learning Evaluation Review and recommends 12 credits for the Cybersecurity Certificate and 8 credits for Data Analytics. Each learning plan runs 60–65 hours plus a 3-hour asynchronous capstone. Announced fall 2024 across the Alabama Community College System, Bay Area Community College Consortium, Colorado Community College System, and Louisiana Community and Technical College System. Co-developed with Jobs for the Future; coursework aligns to CompTIA Security+ topics.
IBM Newsroom, 2024-07-22 · skillsbuild.org/college-students/college-certificates
The flagship community-college announcement describes integration entirely in curriculum terms and names no LMS, API, LTI, SCORM, or Common Cartridge mechanism anywhere. The public certificates page offers exactly one institutional mechanism: an email address. Searching that page's source returns zero occurrences of "canvas", "scorm", "cartridge", or "api".
An LTI 1.3 tool launch renders the provider inside a Canvas iframe. IBM's public terms consent to links only where the link does not "create frames around any page on this Web site or use other techniques that alter in any way the visual presentation." An embedded SkillsBuild launch is therefore not covered by the public terms and requires a written agreement — this is a legal gate, not an engineering one.
"You may not mirror any of the content from this site on another Web site or in any other media." Exporting SkillsBuild content into a SCORM package or Common Cartridge and importing it into a Canvas shell is mirroring. Absent a separate license, the content has to stay on IBM's platform and Canvas can only hold the scaffolding around it.
Credly is named on IBM's own page as "a 3rd party data processor authorized by IBM" for the IBM Digital Badge program; name, email, and badge earned are shared with it. Credly's REST API is well documented and supports listing, issuing, revoking, and bulk search. But every route is rooted at /v1/organizations/{organization_id}/badges and tokens are minted per organization. IBM is the issuing organization, not the college — so a college cannot enumerate its own students' SkillsBuild badges without IBM delegating access. Verifying a single badge whose URL you already hold is anonymous and does work.
Canvas supports LTI 1.3 via Developer Keys with explicit scopes for Assignment and Grade Services (lineitem, score, result.readonly), Names and Role Provisioning (contextmembership.readonly), and Deep Linking. The Content Migrations API imports Common Cartridge and zip packages programmatically, supports selective_import so a curriculum committee can approve subsets, and can place imported items into a named module. Every piece Canvas needs to contribute already exists. The gap is entirely on the SkillsBuild side.
Instructure developer docs · 1EdTech LTI Advantage specifications
An IBM job posting for a technical lead on SkillsBuild for Academia specifies work to "implement and test LTI integrations (LTI 1.3) that connect SkillsBuild offerings with partner LMS systems," and requires familiarity with SCORM, xAPI, and Common Cartridge. This is staffing and intent, not a shipped product — the posting is expired and the original URL is dead. Treat it as the single highest-value question to put to IBM, not as a capability you can plan around.
Under the Department of Education's guidance, metadata about a student's interaction with a service — and even aggregate data — can be PII if a reasonable person in the school community could identify individuals from it. The mitigating control is built into LTI 1.3: the only mandatory identity element at launch is an opaque permanent identifier in the sub claim, and NRPS releases name and email only with explicit platform consent. A pseudonymous launch is achievable by configuration. The harder question is that students create SkillsBuild accounts themselves, which is a different disclosure posture than institution-provisioned accounts and should go to district counsel.
California community colleges recognize five curriculum types in the Program and Course Approval Handbook — degree-applicable credit, nondegree-applicable credit, noncredit, contract education, and fee-based community service. That menu, not the technology, determines how fast a model moves. The highlighted row is my recommendation.
| Model | Speed to launch | Student reach | Curriculum approval | Ease of adoption | Funding | Canvas work |
|---|---|---|---|---|---|---|
| Co-curricular badge pathwayStudent clicks out to SkillsBuild from a Canvas module, student services promotes it, no transcript impact. | ||||||
| Not-for-credit / fee-based community servicesA college-run short course wrapped around the certificate, offered through community education. | ||||||
| Noncredit CDCP Certificate of Completion RecommendedA sequenced noncredit certificate — short-term vocational or workforce preparation — built around the SkillsBuild learning plan. | ||||||
| Credit course adoptionSkillsBuild used as required courseware inside an existing approved course outline of record. | ||||||
| Credit for Prior Learning against the ACE recommendationStudent arrives holding the certificate; the college awards credit under its CPL policy. | ||||||
| New credit Certificate of AchievementA full for-credit certificate built around the SkillsBuild sequence. |
Tier 0 needs nothing from IBM and can run this semester. Tier 1 needs a content license. Tier 2 needs a partner agreement. Building them in this order means the pilot is never blocked waiting on a conversation.
The student reaches SkillsBuild from a Canvas module item — a plain external link, which is squarely within IBM's terms because it does not frame anything. Completion comes back the way IBM's own model already assumes: the student earns a Credly badge and submits its public URL to a Canvas assignment. A small verifier resolves that URL anonymously and confirms the badge, the template, and the earner.
This is deliberately unglamorous. It also works, is legally clean, keeps every credential in the student's own hands, and generates the completion data a program review will need — without asking IBM for anything or sending them a roster.
ExternalUrlcredly.com/badges/{uuid}Most of the work in adopting a third-party certificate is not the content — it is everything a California college has to wrap around it: the course outline of record, student learning outcomes, the distance education addendum, module structure, graded checkpoints, and an accessible syllabus. That scaffolding is identical across every college that adopts the same certificate, and it can be generated once as a Common Cartridge package and imported programmatically.
Canvas's Content Migrations API takes a cartridge, supports selective_import so a curriculum committee approves subsets rather than the whole package, and can drop imported items into a named module. Absent a license, the cartridge carries the college's scaffolding and links out to IBM for the content — mirroring IBM's material into the package is exactly what the terms forbid. That is a question to ask, not a rule to break.
/api/v1/courses/:id/content_migrationscommon_cartridge_importerselective_import=true for committee reviewsettings[insert_into_module_id]This is the integration people picture when they say "put SkillsBuild in Canvas," and it is the one I am proposing to build. An LTI 1.3 tool uses Deep Linking so faculty select individual SkillsBuild learning-plan items into a course, Names and Role Provisioning to read the Canvas roster so nobody hand-enrolls a cohort, and Assignment and Grade Services to write completion back into the Canvas gradebook as a real line item.
Two variants. If IBM ships the tool, the college installs a Developer Key and it is done — and the job posting suggests they have staffed for exactly this. If IBM does not, the college runs the middleware itself: it holds the LTI registration, and gets completion either from a delegated Credly organization scope or an IBM cohort feed. Either variant needs IBM at the table — because of the framing clause, not because of the code.
lineitem · score · contextmembership.readonlyLtiDeepLinkingRequestsub claim, no PII releaseEach of these is currently unresolved in public documentation, and each one changes the architecture. If the focus group can only ask a few things, ask these.
Deep Linking for content selection, Assignment and Grade Services for grade passback, Names and Role Provisioning for rosters — or launch-only single sign-on?
This single answer decides whether a pilot is an integration project or a link-out project.
Through delegated Credly organization scope, a cohort report, or a webhook feed — or must the college fall back on students submitting their own badge URLs?
The entire completion-data architecture forks here, because IBM is the Credly issuing organization and the college is not.
Any embedded launch frames IBM's pages, and any packaged export mirrors IBM's content. Both are prohibited by the terms of use as written.
This is a contract question that has to be settled before an engineering plan means anything.
If yes, Tier 1 becomes a genuine content-delivery path rather than a scaffolding-only wrapper, and progress telemetry becomes available without polling.
Canvas natively imports Common Cartridge; SCORM needs a separate account-level tool and cannot be tested in a sandbox.
Alabama, the Bay Area consortium, Colorado, and Louisiana have been running this since fall 2024. Whatever they built is the real reference architecture.
California already has a participant in BACCC. Their pattern is the fastest thing we could copy.
If accounts are siloed per program, any single-sign-on or account-linking design that assumes one identity per student breaks immediately.
Reported in the FAQ but not firmly established; it constrains every SSO model.
Data and Cyber have a shipped, ACE-reviewed certificate to adopt. AI does not. That asymmetry is worth naming in the room, because it changes what each group should actually be deciding.
The Data Analytics Certificate is built, piloted, and ACE-reviewed. The group's job is adoption mechanics, not design: which model, which department owns it, and how completion is recorded.
The strongest credit case of the three, and the one with an external certification anchor. Same question as Data — adoption model — but with a clearer articulation story into existing CS/CIS programs.
There is no IBM SkillsBuild college certificate in AI equivalent to the Cyber and Data ones — AI content exists in the open catalog, but not as an ACE-reviewed, community-college-designed credential. So this group is not choosing an adoption model. It is specifying a curriculum that does not exist yet.
That is a genuinely different task, and it is the one I can contribute the most to — I have already built a complete AI infrastructure certificate for exactly this gap, with the California regulatory analysis and labor-market justification done.
Most of what a SkillsBuild–Canvas pilot needs, I have already prototyped for other reasons. To be explicit about what these are: they are self-built projects and public repositories, not products any college or district has adopted, procured, or deployed. What they establish is that the technical path is one I have walked before — not that a pilot would be inheriting a running system.
Canvas LMS Administrator for LA Mission College — courses, users, enrollments, and external tool installation. Creating the Developer Key and installing an LTI 1.3 tool is inside my existing role, not a request I have to make of someone else.
LACCD · in placeProject Remedy's Canvas component is an LTI 1.3 tool that scans a Canvas course for WCAG 2.2 AA violations and remediates in place. It is my own project — not installed in any college's Canvas instance and not procured or deployed by LACCD. What it establishes is narrow but real: the LTI 1.3 path into Canvas is one I have implemented end to end, not one I would be attempting for the first time.
Prototype · FastAPI + React · project-remedy.comA command-line tool wrapping the Canvas REST API — 1,042 endpoints — with an offline mirror and cross-resource joins Canvas does not expose. Course shell generation, content migration, module scaffolding, and gradebook writes are scripted and exercised against test instances. Personal tooling, not a district-adopted system.
Prototype · agent-native CLIA desktop tool that opens Canvas IMSCC course exports, remediates them to WCAG 2.2 AA offline, and emits an Accessibility Conformance Report. The cartridge-generation path in Tier 1 would reuse machinery that already exists in prototype form rather than starting from a blank page.
Prototype · Tauri + PythonAn eight-course, 540-hour AI Infrastructure and Architecture Certificate of Completion, with a self-authored compliance analysis against PCAH 8th edition, Title 5, and the Education Code sections governing CDCP enhanced funding, plus EDD labor-market justification. It has not been through local curriculum approval or state submission. Treat it as a worked template for how a SkillsBuild certificate could be chaptered in California, not as an approved program.
Draft · published at aiinfra.johnnyphung.comA frozen, reproducible evidence release mapping AI exposure, wages, and openings across 64 occupations in Los Angeles against a 41-college curriculum-response sample — built specifically for CTE deans and curriculum committees who have to defend a program proposal. It answers the "where is the regional need" question the breakouts will run into.
Independent analysis · benchmark.johnnyphung.comA prototype AI-assisted curriculum management system in which faculty would author course outlines of record and SLOs through role-based approval, grounded in PCAH and Title 5 — plus a tool that converts documents into Canvas-importable QTI assessments with LaTeX and scientific notation support. Public repositories, not systems any college has adopted.
Prototype · Next.js + FastAPISelected as an AI ChangeMaker in the California Community Colleges Chancellor's Office statewide AI initiative, and a member of the LACCD District Academic Senate AI Taskforce — which is where a statewide deployment recommendation actually has to travel to become policy.
CCCCO · 116 colleges, 2.1M studentsNothing in IBM's public documentation offers one, and the terms of use actively block the two obvious workarounds. Tier 0 delivers students this semester and produces the completion data a program review needs. Start there and let the agreement conversation run in parallel.
It is the only model that is free to the student, open-entry, and funded at credit parity. For a statewide deployment recommendation across 116 colleges, that combination is what makes reach and sustainability compatible.
Every college adopting the same certificate rebuilds the same course outline, SLOs, DE addendum, and module structure. Generating that scaffolding once as a shareable cartridge is the highest-leverage thing this group could produce — and it needs no permission from IBM.
Alabama, BACCC, Colorado, and Louisiana have been running this since fall 2024. Whatever they actually implemented is worth more than any architecture diagram, including this one.
There is no ACE-reviewed AI college certificate to adopt. The AI group's real output is a specification for one — and California, with an existing statewide AI initiative, is well positioned to co-design it rather than wait for it.
Students self-creating vendor accounts is a different disclosure posture than institution-provisioned ones. LTI 1.3 can launch pseudonymously by configuration, but that has to be a decided design constraint, not an afterthought.