Contract-review training
Practice spotting missing indemnification clauses, ambiguous IP terms, hidden auto-renewals. Trained on contracts that look like the ones associates actually see.
Coming soon
Turn any document into a spot-the-issue drill with Claude or ChatGPT.
Build review-training activities in plain language. Learners spot issues in contracts, charts, code snippets, ad copy, or clinical notes. AI grades against your rubric and credits learners for the issues they find and the false positives they avoid.
What is it?
What is a document-review training activity?
A training activity where the learner reads a document, contract, chart, code snippet, ad copy, lab note, and marks the issues. Both true findings and false-positive flags are graded, so the learner is trained to be precise as well as thorough.
-
Highlight-and-annotate UI on real-looking documents, contracts, charts, code, copy, clinical notes -
Issues seeded by the AI based on rubrics you provide, with control over difficulty and density -
Per-issue feedback, learners see what they missed, what they correctly caught, and what they wrongly flagged -
False-positive penalty configurable to match real-world consequence weighting -
Exports as a SCORM 1.2 package for Cornerstone, Moodle, Canvas, TalentLMS, and every LMS -
Same bundle runs standalone in any browser for self-study without an LMS
How it works
How to build a document-review activity
in four steps.
From a plain-English description of what good and bad look like, to a SCORM-ready review exercise. No PDF annotation tool integration required.
01
Provide the source document
Paste a contract, code snippet, chart, ad copy, or clinical note. The tool renders it in a review-friendly UI with highlight and annotation tools.
02
Define the issue types
Name what counts as an issue, missing indemnification clause, off-by-one error, vague claim, missing CTA. The AI seeds them at the density and difficulty you set.
03
Tune in chat
"Add a subtle indemnification gap." "Throw in two false-flag items that look risky but aren't." Iterate by talking with your coding agent, not by editing markup.
04
Export as a SCORM 1.2 package
One command produces a SCORM 1.2 zip. Upload to any SCORM LMS. Precision, recall, and total score post back to the gradebook automatically.
Use cases by role
Document-review training
for every careful-reading job.
The same toolchain powers contract-review training, code-review onboarding, ad-copy QA, clinical-note auditing, and editorial fact-checking drills.
Code-review onboarding
New engineers practice spotting bugs, security issues, and style violations on realistic diffs before reviewing real PRs. Calibrate the issue density to the team's standards.
Ad-copy & claims review
Train reviewers to catch unsupported claims, regulatory red flags, and tone problems in ad copy and product pages. Especially useful in regulated industries.
Clinical-note auditing
Practice spotting documentation errors, coding mismatches, and clinical-quality issues in chart notes, the kind of review training payers and quality teams need.
Fact-check drills
Train editorial staff to flag unverified claims, missing sourcing, and quote-accuracy issues. Realistic articles with seeded errors at the density you choose.
Financial-statement review
Practice spotting classification errors, suspicious year-over-year movements, and disclosure gaps in realistic statements.
Precision- and recall-graded review
Every flagged passage graded for accuracy, both finds and false flags.
When the review ends, the AI scores recall (issues caught) and precision (issues correctly identified vs falsely flagged) separately. Total score weights both, so learners can't game it by flagging everything.
Update the rubric, issue density, or precision-vs-recall weighting any time. Re-export the SCORM package and the next attempt grades against the new rules.
Precision-and-recall scoring
Every issue caught counts toward recall. Every false-positive flag counts against precision. Total score weights both, so learners can't game it by flagging everything.
AI-seeded issues
The AI inserts issues into your source document based on the rubric and difficulty you set. You review the seeded issues before publishing.
Layered difficulty
Start with obvious issues and progressively introduce subtler ones. The learner sees how their precision degrades as issues get sneakier.
Session complete
Your Debrief
9/10
★★★★★★★★★☆
Total 9 / 10
Caught the subtle indemnification gap in section 7.2 that the prior cohort missed in 60% of attempts.
Did not flag the assignment-with-consent clause as a problem, correctly recognizing it as standard.
Maintained precision above 80% as issue subtlety increased through the second half of the document.
Plain-English authoring
Author document-review activities in plain English. Paste a doc, get a SCORM exercise.
Add issue types, change density, raise the difficulty, or adjust the precision-vs-recall weight, just describe the change in your own words. Your coding agent reseeds the document and re-exports the SCORM package without you ever editing markup.
When you're happy, export as a SCORM 1.2 package for your LMS, or use the same HTML bundle standalone.
Add two subtle false-flag passages that look risky but are actually standard.
Added 2 false-flag passages: a typical force-majeure clause and a standard governing-law provision. Precision rubric updated.
Make precision count for 60% of total instead of 40%.
Reweighted: precision
0.6, recall 0.4. Total grader recompiled.Export as SCORM.
✓ document-review.zip ready (SCORM 1.2, 14 KB)
Review Library
Start from real review
assignments.
Paste any prompt into your coding agent to get a complete document-review activity as a single self-contained HTML file. Adjust source doc, issue rubric, or weights in plain language.
Legal
SaaS MSA review
An associate-level review activity on a 6-page SaaS MSA with 8 seeded issues and 2 false-flag clauses. Score weights precision over recall.
Prompt
Using /document-review, build a SaaS MSA review activity with 8 issues and 2 false flags. Score with precision-weighted rubric.
Engineering
Backend PR review
A 200-line backend PR with 6 seeded issues (one security, two correctness, three style) and 3 false-flag passages that look risky but aren't.
Prompt
Using /document-review, build a backend PR review activity with 6 issues across security, correctness, and style. Score precision and recall separately.
Marketing
Pharma ad-copy review
An ad-copy review for a regulated pharma product with 7 seeded issues (unsupported claim, off-label hint, fair-balance gap) and 2 false flags.
Prompt
Using /document-review, build a pharma ad-copy review activity with 7 issues and 2 false flags. Score on fair-balance, on-label, and substantiation.
Clinical
Inpatient note audit
A 1-page inpatient progress note with 5 seeded documentation/coding issues and 2 false flags. Grader weights recall on coding errors.
Prompt
Using /document-review, build an inpatient note audit with 5 issues across documentation and coding. Score recall-weighted on coding errors.
Editorial
News-article fact check
A 600-word news article with 6 seeded fact issues (unsourced claim, quote attribution error, date inconsistency) and 2 false flags.
Prompt
Using /document-review, build a 600-word fact-check activity with 6 issues across sourcing and accuracy. Score on attribution and verification.
Finance
10-K disclosure review
A 2-page excerpt from a 10-K with 5 seeded issues (classification, YoY anomaly, disclosure gap) and 3 false flags. Aimed at audit-staff training.
Prompt
Using /document-review, build a 10-K disclosure review with 5 issues and 3 false flags. Score on classification, YoY anomaly, and disclosure completeness.
SCORM & LMS
SCORM 1.2 document-review activities for any LMS.
Every document-review activity exports as a standards-compliant SCORM 1.2 package. Upload the zip, assign it like any other course activity, and precision, recall, and total scores flow back to the gradebook automatically.
-
Single SCORM 1.2 zip, upload to your LMS, no integration work, API keys, or custom JavaScript -
Completion, score, and time-on-task post back via cmi.core.score.rawandcmi.core.lesson_status -
Tested with Cornerstone OnDemand, Moodle, Canvas LMS, TalentLMS, Docebo, Brightspace, Absorb, 360Learning, SuccessFactors, and Workday Learning -
Same self-contained HTML bundle runs standalone on a public page, an intranet, or as an embedded iframe -
Document-agnostic UI works for contracts, code, charts, ad copy, and clinical notes -
Grading runs inside the bundle, no middleware, no data warehouse, no analytics SDK required
FAQ
Frequently asked questions about
document review.
Authoring, precision and recall, LMS compatibility, document formats, model choice, answered directly.
A document-review training activity is an interactive exercise where the learner reads a document and marks the issues. Both correctly caught issues (true positives) and incorrectly flagged passages (false positives) are graded, training the learner to be precise as well as thorough.
Anything text-based or structured, contracts, code snippets, charts, ad copy, clinical notes, financial statements, lab reports, editorial articles. The activity renders the document in a review-friendly UI and adapts to the document type.
The AI inserts issues into your source document based on the rubric and difficulty you specify. You review and approve the seeded issues before publishing the activity.
Every issue caught contributes to recall. Every passage incorrectly flagged contributes to a precision penalty. Total score is a weighted combination you control, so learners are trained to be accurate, not just thorough.
Yes. Every activity exports as a SCORM 1.2 package, which works in Cornerstone, Moodle, Canvas, TalentLMS, Docebo, Brightspace, and every other SCORM-compliant LMS. Precision, recall, and total score post back automatically.
Yes. Paste your own contracts, code, or copy and let the AI seed practice issues into them. This is how teams produce review training that looks exactly like the documents reviewers will see at work.
The bundles run on whichever frontier model you configure, Claude, GPT, Gemini, or a self-hosted model. The model is swappable; the rubric, seeded issues, and grading logic are stored in plain text and are model-agnostic.
Document Review is in active development. It will install via your coding agent with a single command, the same way the existing role-play skill works today. Add your email below and we'll let you know the day it ships.