Rfp Response Content Finder

Content Finder Skill

Firm Profile (REQUIRED — fill in before use)

These values define who your firm is and how it writes. See SETUP.md.

• FIRM NAME: [FIRM NAME]
• WHAT THE FIRM DOES: [one sentence]
• ROLE BOUNDARY: [what the firm will NOT claim to do — leave blank if none]
• VENDOR STANCE: [vendor-neutral OR resells/implements specific products]
• SPECIALIZATION: [domains the firm credibly claims]
• VOICE: [3–5 adjectives]

Purpose

For each row in the RFP Tracker, classify it, route it to the most relevant content libraries, and populate four columns: Available Content (a short plain-language summary for human review in list view) and up to three Source hyperlink columns pointing to the matched documents.

Source links are the drafting input for Phase 3. The Content Generation agent follows them to retrieve and read the full documents before drafting. The Available Content summary is for human review only — it is not sufficient for drafting.

Prerequisites

• The RFP Tracker List must be populated (output of RFP Extraction).
• The list must have these columns configured:
  • Available Content — Multiple lines of text
  • Source 1, Source 2, Source 3 — Hyperlink type (each stores one URL with display text)
• Read access to the firm's content libraries:
  • Capability Library — methodology, capability statements, case-study abstracts
  • Content Asset Library — slides, FAQs, proposal fragments, boilerplate
  • Past Proposals Library — prior submitted proposals
  • Standard Positions — reusable positions on IP & data rights, pricing, exceptions & assumptions
• Write access to the RFP Tracker list to update all four columns.

Library names above are the SETUP.md defaults. If you renamed them, use your names consistently here and in the routing map.

Step 1 — Load the RFP Tracker List

1. Retrieve all items from the RFP Tracker list.
2. For each item, read Item, Category, Mandatory, and the four output columns — skip rows where all four are already populated unless the user requests a full re-run.
3. Build a working queue. Place Mandatory = Yes rows at the front.

Step 2 — Pre-screen: Classify Each Row

Classify every queued row as Administrative or Substantive before searching — this determines whether a library search is needed and which libraries.

Administrative Requirements

Pass/fail legal, corporate, or regulatory qualifiers. The correct response is a legal attestation, standard corporate certification, or a direct lift of identity data — not a content match from the methodology libraries.

Indicators: legal authority to bind the firm; corporate registration, entity status, jurisdiction, tax IDs; financial stability, bonding, insurance; conflict-of-interest/independence; signatory authority/notarization; diversity/ socioeconomic certifications.

Action: no methodology-library match needed. Write a standard attestation note pointing to where the corporate/legal data lives and, where appropriate, the team that confirms response language.

Available Content: [ADMINISTRATIVE QUALIFIER] Pass/fail legal or corporate
eligibility requirement. Respond with standard corporate certification,
attestation, or legal declaration. Identity/boilerplate data — if needed —
comes from the firm's corporate boilerplate document.

Source 1: [Display: Corporate Boilerplate — Content Asset Library]
          [URL: link to the firm's corporate boilerplate document]
Source 2: [Display: Legal/Contracts team — confirm response language]
          [URL: leave blank or link to internal template if available]
Source 3: [leave blank]

Substantive Requirements

All other rows — approach, methodology, team, references, pricing, governance, change management, IP/data rights, exceptions, technical capability. Apply Steps 3 and 4.

Step 3 — Route, Then Index the Document Libraries

For substantive rows only. Route first, then index — searching every library for every row wastes budget and dilutes relevance.

Library Routing Map

Adapt the topic rows to your firm's actual SPECIALIZATION. The routing structure matters more than the specific keywords.

Indexing

For each routed library:

1. Retrieve documents.
2. Extract: title, filename, library, full text (or, for long documents, key sections — executive summary, approach, methodology, differentiators, team profiles).
3. Log any documents that cannot be read (scanned, protected, corrupt) and continue — do not halt.

If a routed library has 50+ documents and the requirement is broad, ask the user to confirm the RFP's primary solution domain so results can be pre-filtered by metadata tags before scoring.

Step 4 — Score and Rank Content

For each substantive row, score documents from the routed libraries.