| name | contract-snapshot |
| title | Contract Snapshot |
| description | Use when the user wants to compare the same handful of terms across N contracts side-by-side in a grid — what is the term, survival period, carveouts, and governing law in each of these 5 NDAs? Returns a row-per-document × column-per-question grid with citations per cell. Reference skill for the M3-C output_format - table mode; intended as a starting point for operators to fork and tune for their own contract types. |
| author | LegalQuants |
| author_url | https://github.com/LegalQuants/lq-ai/tree/main/skills/contract-snapshot |
| license | Apache-2.0 |
| version | 0.1.0 |
| execution_mode | open |
| jurisdiction | general |
| practice | contracts |
| language | en |
Contract Snapshot
A reference skill for the M3-C output_format: table mode. Produces a side-by-side grid of the same questions across N contracts — the in-house lawyer's "compare clauses across N agreements" workflow. Each cell carries a citation back to the source document, and failed extractions render as not found rather than confidently-wrong text.
When this skill applies
Apply when the user wants to compare a small number of well-defined questions across a corpus of similar contracts:
- "What is the term, survival, and governing law across these 5 NDAs we're tracking?"
- "Pull out the payment terms, IP ownership, and termination triggers across these 10 MSAs."
- "For my Q3 portfolio review, I need a grid of these 30 vendor contracts' liability caps."
Do not apply this skill to:
- Single-document review — use the appropriate document-specific skill (
nda-review, msa-review-saas, etc.).
- Free-form chat against contracts — that's the regular Chat surface.
- Tasks where the questions aren't well-defined upfront — the column queries must be specific enough to extract from each row's source document; vague queries produce poor cells.
Inputs
The skill takes a set of documents (selected via the Tabular Review UI from a Knowledge Base, a Project, or a free file selection). The four columns above run as Citation Engine-grounded extractions against each document.
To adapt this skill for a different contract type (e.g., MSAs), fork the skill and rewrite the four column queries. Keep them short, specific, and quote-asking — the Citation Engine works best when the model is encouraged to quote rather than paraphrase.
Per-column overrides
This skill demonstrates the two per-column overrides M3-C1 supports:
ensemble_verification: true on the Survival column. Survival is the load-bearing economic term in confidentiality agreements (a 3-year confidentiality term with a 10-year survival is very different from one with no survival), so cells in this column run through Stage 4 of the Citation Engine cascade — three judges debating whether the cell value is faithful to its citation. Higher cost, higher confidence.minimum_inference_tier: 3 on the Governing Law column. The skill-level floor is Tier 2 (commercial inference). Governing-law extraction is the column most likely to surface counterintuitive answers (e.g., a contract drafted under California law but with a Delaware forum-selection clause); routing this column to Tier 3+ avoids the cheapest models' tendency to collapse the two into one answer.
