Bid Contract

Bid-contract — a meta-prompt that emits deal-specific paper

This skill is not a document outline. It is a typed clause system: a shared clause library plus a floor model and pricing models, from which each run assembles the specific paper one deal needs. The library is machinery reused across every client; only the bound slot values are deal-specific. Read ../forge/references/evidence-rules.md and ../forge/references/meta-meta-map.md before generating. Plain language throughout — a buyer's operations manager must understand every clause without counsel.

Typed inputs (bind at intake; UNKNOWN is legal — never invent)

SYNTHESIS:                 path            # 01-research/SYNTHESIS.md — the ONLY number source (one-directional)
DEAL_SIZE_BAND:            {small|mid|large}   # sets safety factor + clause depth + which gauntlet to expect
BILLABLE_UNIT_CANDIDATES:  list[str]       # candidate atomic units from SYNTHESIS (booked appt, shipped skill, trained seat…)
JURISDICTION:              str|UNKNOWN      # governs law, liability cap, data regime — UNKNOWN never silently defaulted
BUYER_PROCUREMENT_MATURITY:{low|medium|high|UNKNOWN}  # which procurement gauntlet to expect
CHANGE_APPETITE:           {augment|transform|explore}|UNKNOWN  # biases pricing-shape recommendation + tone

Also consumed from SYNTHESIS: the TELL ↔ SCOPE ↔ PRICE map, critical-path gates, cost-floor inputs, RELATIONSHIP_WARMTH, BUDGET_SIGNAL, and proof status.

Typed outputs

pricing-floor.md      # INTERNAL, never shipped — the computed floor + sensitivity (also .xlsx when xlsx installed)
proposal-draft.md     # client-facing — situation, findings, sign-here scope, definitions/SLA, timeline, 3 shapes, TBDs
contract-skeleton.md  # client-facing draft — assembled clauses with bound slots, attorney-review caveat
[proposal.docx]       # optional, when the docx skill is installed

Procedure — Frame → Do → Review → Capture

Frame. Read SYNTHESIS. Bind the typed inputs. Confirm every number you will use exists in SYNTHESIS (never introduce a number inside an artifact). List the gates and the tell↔scope↔price map.

Do.

1. Compute the floor → pricing-floor.md per references/floor-math-worksheet.md. Every cost line sourced or EST-flagged with basis. Run the ±50% sensitivity band. This is the number every price must clear.
2. Define the billable unit. Pick from BILLABLE_UNIT_CANDIDATES the unit with the tightest checkable predicates and a system of record; draft it via references/clause-library.md family 4 with a dispute window (family 5).
3. Select pricing shapes via the decision tree in references/pricing-shapes.md over proof status, CHANGE_APPETITE, RELATIONSHIP_WARMTH, BUDGET_SIGNAL. Recommend one; always offer all three. Every shape priced ≥ floor × SAFETY_FACTOR.
4. Assemble the paper from the clause library, binding each {{slot}} from SYNTHESIS/intake. Select clause families per their when-to-omit rules and the chosen shape. Render the sign-here scope schedule as the contract's backbone. UNKNOWN slots render as [[TBD: one-sentence question]] — strictly for jurisdiction-dependent terms (governing law, liability cap, data regime).
5. Anticipate procurement per references/procurement-checklist.md: fold gate-clearing costs into the floor, flag any red-flag clause on the buyer's paper, set the timeline against the procurement gate.

Review. Run the predicates below. Any failure ⇒ fix the input and regenerate — never hand-patch a clause in a shipped draft (functoriality; ../forge/references/meta-meta-map.md).

Capture. Mint atoms to FACTORY/atoms.md: clause outcomes (which clause positions held or were conceded), pricing-shape win rates, procurement patterns per buyer type. Log human corrections and their trace to FACTORY/feedback-loop.md (see Feedback tracing).

Artifact specs + acceptance checklists

pricing-floor.md — table of cost lines (delivery infra, labor role×hours, gate-clearing, tooling, payment fees) + overhead + safety factor + minimum price + sensitivity band. ☐ every line sourced or EST-flagged ☐ overhead allocated ☐ sensitivity computed ☐ minimum price stated.

proposal-draft.md — Situation (their context, their language) · What we found (tells with citations, respectful version) · Scope — the sign-here table (tell ↔ scope ↔ price, unresolved tells explicitly deferred) · Definitions + SLA (billable unit defined contractually; response times; reporting cadence = the dashboard) · Timeline keyed to gates (each gate named, owner, honest duration; no result promises before gates clear) · Three pricing shapes (recommended first; each: cost, inclusions, when it fits THEM) · TBDs (every surviving UNKNOWN as a one-sentence question). ☐ bijection holds ☐ all prices clear the floor ☐ timeline references only cited SYNTHESIS gates ☐ no pre-proof volume/outcome promise ☐ attorney-review caveat present.

contract-skeleton.md — assembled clauses: parties (entity UNCONFIRMED until verified) · term + pilot conversion · scope schedule · billable unit + dispute window (verbatim-consistent with proposal) · payment · data handling + ownership · IP of built systems · SLA + reporting · termination + wind-down · liability cap · non-solicit · publicity — each selected per its when-to-omit rule. ☐ every clause plain-language ☐ jurisdiction-dependent slots bound or flagged TBD ☐ dispute window identical across all clauses that reference it ☐ attorney-review caveat present.

Scale-native design

The clause library, pricing models, floor worksheet, and pre-staged procurement pack are shared across all clients (edit once, every run benefits). Deal-specific values are isolated per run in prospects/<slug>/03-bid/. To generate paper for a batch of prospects, run the procedure per slug in parallel after each SYNTHESIS is vetted — each bid is an independent projection of one SYNTHESIS, so they fan out without collision (../forge/references/meta-meta-map.md, monoidal composition).

Predicates (checkable — all must pass before shipping)

• Floor cleared — every shipped price ≥ floor × SAFETY_FACTOR; if not, scope changed, not floor.
• Bijection complete — no orphan scope, no orphan tell, no orphan price line.
• No pre-proof commitments — zero volume/outcome guarantees; the pilot carries calibration.
• TBDs explicit — every surviving UNKNOWN is a flagged one-sentence question, never a silent default; jurisdiction-dependent terms flagged when JURISDICTION is UNKNOWN.
• Numbers sourced — every figure traces to SYNTHESIS; Stage-3b corrections applied; entity names UNCONFIRMED until verified.
• Consistency — dispute window, billable unit, and prices are identical across proposal and contract (they factor through SYNTHESIS, never copied artifact-to-artifact).

Core invariants

Sourced-or-flagged numbers · UNCONFIRMED entity flags until human-verified · plain language in all client-facing paper · every emission carries the attorney-review caveat (this is a negotiation-ready draft, not legal advice) · regenerate on refined input, never hand-patch a shipped clause.

DONE + refusal conditions

DONE when: all three files exist · floor computed with sensitivity · all prices clear the floor · billable unit contractually defined with a dispute window · bijection holds · TBDs listed · procurement gauntlet anticipated · attorney caveat present · ≥1 atom minted.

Refuse (surface the reason, do not silently comply): to price below the computed floor — descope instead; to write any volume/outcome commitment before proof — offer a pilot that carries calibration; to silently default a term when JURISDICTION is UNKNOWN — flag it as a TBD for the client's attorney. Also refuse to assign background/general-purpose IP or accept uncapped liability (references/procurement-checklist.md red flags).

Feedback tracing

After human feedback, trace each correction to the exact source that allowed it — a clause family, a typed slot, a pricing-shape branch, or a floor line — then edit that shared reference (never the run's output), bump the version, and re-test the edit against 2–3 other deal types so the fix generalizes. Log to FACTORY/feedback-loop.md; mint the reusable lesson to FACTORY/atoms.md.

References index

• references/clause-library.md — the 13 typed clause families: purpose, typed slots + legal-safe defaults, plain-language pattern with {{slots}}, negotiation range (principles), red lines, when-to-omit. Read when assembling any contract paper.
• references/pricing-shapes.md — pilot/hybrid, performance-max, classic retainer as typed models with formula skeletons, buyer profiles, failure modes, and the recommendation decision tree. Read when selecting and pricing shapes.
• references/floor-math-worksheet.md — the computable floor: cost lines, overhead, margin of safety, ±50% sensitivity, output table, and the xlsx hook. Read when computing pricing-floor.md.
• references/procurement-checklist.md — surviving buyer procurement at scale: the gauntlet, response SLAs, the pre-staged document pack, and red flags that reprice or kill deals. Read when the buyer has a procurement function or issues their own paper.

Core-pack note (read-only, not edited by this skill)

factory-schema.md §5 could add procurement to its named critical-path gate list explicitly, and its DONE predicate could reference the dispute-window doctrine now formalized here — noted for the forge maintainer, not changed by this skill.