alterlab-syllabus-ai-policy — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited alterlab-syllabus-ai-policy (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
The focused add-on that turns "what's my AI policy?" into a concrete, paste-ready syllabus statement. It does one thing well: given a course and its graded tasks, it assigns each task an explicit permitted / restricted / prohibited tier, writes the matching disclosure and attribution clause, and binds the whole statement to the institution's own academic-integrity code — so the policy is enforceable, not aspirational. It deliberately does not author the rest of the syllabus, learning outcomes, or rubrics; that is alterlab-teaching-design.
Use it when the request is about the AI-use rules of a course or assignment:
Draft an AI-use policy for my syllabus.
Write a course statement on ChatGPT / generative AI for students.
I need an academic-integrity clause that covers AI tools.
Give me per-assignment rules: where can students use AI, where not?
How should students disclose and cite AI they used in an essay?
Make my AI policy consistent with our university's integrity code.→ Gather the course context (level, discipline, the list of graded tasks, the institution's integrity-code reference), assign each task a tier, then run scripts/policy_builder.py to emit the statement and scripts/policy_lint.py to catch contradictions before you hand it back.
This skill is a narrow add-on. Route adjacent asks to the right sibling:
| The ask is really about… | Route to | Why not here |
|---|---|---|
| Designing the whole course / syllabus, learning outcomes, rubrics, lesson plans, backward design | alterlab-teaching-design | Owns full course/backward design; this skill only writes the AI-policy section |
| Ethics of using an AI tool on human-subjects data (IRB, consent, de-identification) | alterlab-research-ethics | Research-ethics / IRB territory, not a teaching policy |
| Whether a student submission was AI-generated; running a detector | alterlab-teaching-design (assessment) | This skill writes policy; it does not adjudicate or detect individual cases |
| Verifying that citations a student or author produced actually exist | alterlab-citation-verifier | Citation existence-checking, not policy drafting |
| Turkish-system integrity/ethics process (ÜAK, YÖK etik kurul) | alterlab-tr-research-ethics | Turkey-specific ethics workflow, parameterized differently |
| Institutional accreditation / assurance-of-learning reporting | alterlab-accreditation-aol | Program-level AoL, not a course AI clause |
If the request mixes "design my course" and "write my AI policy", do the AI policy here and hand the rest to alterlab-teaching-design.
Every graded task is assigned exactly one tier. The three-tier shape mirrors the Cornell University faculty-committee framework (prohibit / allow-with-attribution / encourage) — a published, citable model — but the wording is yours and is bound to your institution's integrity code, never invented.
| Tier | Meaning | Student obligation |
|---|---|---|
| Prohibited | No generative-AI use; the task measures a skill AI would substitute for | None permitted; use is an integrity violation under the cited code |
| Restricted | AI permitted for named sub-tasks only (e.g. brainstorming, grammar), not for the assessed deliverable | Disclose what tool was used, for what, and how — see the disclosure clause |
| Permitted | AI use is allowed and may be encouraged (e.g. as a tutor, for accessibility) | Disclose and attribute per the citation template; remain responsible for accuracy |
Default-deny when unstated. If the course gives no tier for a task, the statement says the task is Prohibited until the instructor decides — never silently "anything goes". A vague "students may use AI responsibly" line is the single most common failure and policy_lint.py flags it.
See references/tier_framework.md for the full decision tree (which tier fits which assessment type), worked per-discipline examples, and the verbatim Cornell tier definitions this is modeled on.
A tier alone is not a policy. Restricted and Permitted tasks require students to disclose and cite AI use, and the statement must hand them an exact citation format — not "cite it appropriately". Use the documentation standard the course already uses:
company, not the tool. Cite the specific chat when it generated ideas, content, or information: Company. (Year, Month Day). Title of chat [Generative AI chat]. Tool/model. URL-of-shared-chat; cite the tool for general uses such as editing your own writing: OpenAI. (2025). ChatGPT (GPT-5) [Large language model]. https://chatgpt.com/. The chat title and the tool name are italicized. In-text (OpenAI, 2025). (This replaces the 2023 "Mar 14 version" example.)
template of core elements, with the tool as the container, not the author, and the model named in the version element: "<Description of prompt>" prompt. ChatGPT, model <model>, OpenAI, <date>, <shareable chat URL>. Use the tool's general URL only when it offers no share link. MLA also asks writers to acknowledge functional uses (editing, translation) in a note.
references/disclosure_and_citation.md carries both templates with examples, a Chicago-style note option, and a ready-made student "AI-use declaration" block the statement can append to each submission.
Three clauses every statement should carry, all parameterized — never asserted as fact about a specific tool:
institutional code (e.g. "violations are handled under <CODE_REF>"). This is what makes the policy enforceable; leave <CODE_REF> as a fill-in if the user has not supplied it, and say so.
and that not all students have paid-tier access, so required AI use should be free-tier-achievable or provided. Do not claim a specific tool is accessible/compliant unless the user supplies that fact.
detector", flag that detector outputs are probabilistic and should not be the sole basis of an integrity finding; the policy states process, it does not adjudicate. (Adjudicating an individual case is out of scope — see the routing table.)
Detail, sample wording, and the equity checklist live in references/integrity_accessibility.md.
uv run python skills/faculty-life/alterlab-syllabus-ai-policy/scripts/policy_builder.py \
--course "PSY 201 Research Methods" --level undergraduate \
--doc-standard apa \
--integrity-code "IEU Student Disciplinary Regulation" \
--task "Literature review essay=restricted:brainstorming and outlining only" \
--task "In-class exam=prohibited" \
--task "Data-analysis report=permitted:as a coding tutor" \
--out ai_policy.md--task "<name>=<tier>[:<scope/notes>]" — repeatable; tier ∈prohibited|restricted|permitted. A task with no tier is emitted as prohibited with a visible "instructor to confirm" note.
--doc-standard apa|mla|chicago|none selects the citation template block.--integrity-code "<ref>" is inserted verbatim; omit it and the output keeps a<CODE_REF> placeholder plus a warning.
--out to print the Markdown statement to stdout.The script is stdlib-only (no network, no keys): it fills local templates and never invents an integrity code, a tool capability, or a citation URL.
uv run python skills/faculty-life/alterlab-syllabus-ai-policy/scripts/policy_lint.py ai_policy.mdpolicy_lint.py flags the common failure modes: a vague catch-all ("use AI responsibly") with no per-task tier, a Restricted/Permitted tier with no disclosure clause, a missing or placeholder integrity-code reference, a citation instruction with no concrete template, and "prohibited" language that contradicts a "permitted" line for the same task. It exits non-zero on any error-level finding so it can gate a CI or a pre-handoff check.
Present the statement, the lint verdict, and every unresolved placeholder (<CODE_REF>, unconfirmed tool claims). If the user also wants the surrounding syllabus, outcomes, or a rubric, hand off to alterlab-teaching-design rather than improvising them here.
to Prohibited (not "responsible use")?
concrete citation template (not "cite appropriately")?
placeholder — never an invented regulation name?
accessible unless the user gave you that fact?
policy_lint.py pass (no error findings)?references/tier_framework.md — the prohibit/restrict/permit decision tree,per-assessment-type guidance, worked examples, and the verbatim Cornell faculty-committee tier definitions this models.
references/disclosure_and_citation.md — APA 7 and MLA 9 AI-citationtemplates as revised in 2025, a Chicago note option, plus a student AI-use declaration block.
references/integrity_accessibility.md — integrity-binding wording,accessibility/equity checklist, and the detector-as-evidence caution.
Intelligence in Education* — prohibit / allow-with-attribution / encourage course-policy framework. teaching.cornell.edu.
(September 2025) — cite a specific chat [Generative AI chat] or the tool [Large language model], company as author; supersedes the 2023 How to cite ChatGPT example. apastyle.apa.org (templates cross-checked against the UBC and Oregon State University library guides, 2026-09-23).
Revised)* (August 2025) — tool as container, model in the version element, shareable chat URL preferred. style.mla.org.
Part of the AlterLab Academic Skills suite.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.