Playbook Easy Extract

Extract the negotiated positions from a single contract — the clauses where the parties take a substantive stance on a recurring contract issue — and emit a structured list the Easy Playbook generation pipeline can cluster.

When this skill applies

This skill is internal to the Easy Playbook generation pipeline (M3-A6) — it is not a user-facing skill. The pipeline calls it once per uploaded corpus document; the per-document output is then clustered across the corpus to detect modal positions, and the result is assembled into a draft playbook the user-attorney reviews and edits.

Apply when given a single contract text and asked to enumerate the negotiated positions it takes. Do not apply for contract review, redlining, or any user-facing task — those are the job of the per-contract-type skills (nda-review, msa-review-saas, etc.).

Workflow

Read the contract end-to-end before emitting any output. Identify every clause that:

1. Takes a substantive position on a recurring contract issue. Recurring issues are those that appear across most contracts of the same family: confidentiality definition, term, limitation of liability, indemnification, payment terms, governing law, etc. A clause that imposes an obligation, defines a key term, allocates risk, or sets a threshold is "substantive."
2. Is recognizable as a position (not boilerplate). Skip pure mechanical clauses (notices, severability, integration) — they don't represent a negotiated stance worth clustering. Include them only when the contract takes an unusual position on them.
3. Is contained in identifiable, contiguous text. Don't emit overlapping or fragmentary spans — pick the section/sentence that captures the position.

For each identified clause, emit one entry. The same contract typically yields 5–20 entries depending on length and complexity.