Skip to content

BdM-15/capture-insights / skills/rfp-reverse-engineer

rfp-reverse-engineer

Verified

Reverse-engineers a federal RFP we received — given the SOW/PWS and evaluation criteria already in the Theseus KG, reconstructs the CO's hidden decision tree (upstream `sow-pws-builder` 6 scope blocks + 3 intake answers), surfaces hot buttons, ghost language, discriminator hooks, missing-section signals, and CPFF-form / Section-5 / QASP / Key-Personnel traps. USE WHEN the user asks "what scope decisions did the CO already make?", "reverse engineer this RFP", "what hot buttons are hiding in this PWS?", "where are the discriminator hooks?", "did they pick CPFF completion or term form?", "anything suspiciously missing?", or any variant of decoding CO intent. Pulls `requirement`, `deliverable`, `proposal_instruction`, `evaluation_factor`, `clause`, `performance_standard` from the active workspace KG and emits a JSON envelope feeding `proposal-generator`. DO NOT USE FOR proposal prose (`proposal-generator`), pricing (`price-to-win`), clause audit (`compliance-auditor`), or sub SOW (`subcontractor-sow-builder`).

references/narrative_template.md

View on GitHub
# RFP Reverse-Engineer — Capture-Team Narrative Template

**Purpose:** Lower the degree of freedom in step 10 of `SKILL.md`. The model fills in this exact bullet skeleton. Every bullet is one of two forms:

- **Class A bullet** (factual): ends with `[see <jsonpath>.source_chunk_ids]` (cheapest), `[chunk-xxxx]`, `[entity: <Name>]`, or `[per references/<file>.md]`.
- **Class B bullet** (judgment): begins with one of the visible-judgment markers — `**Our read:**`, `**In our capture experience,**`, `**Likely …**`, `**Classic … pattern —**`, `**Rule of thumb:**`.

**No paragraphs.** **No bullet without one of those tail-anchors or lead-markers.** If you cannot honestly anchor or honestly frame as judgment, drop the bullet.

The `<jsonpath>` form is the cheapest and the audit trusts it because the JSON envelope is on disk with `source_chunk_ids` arrays on every entry. Use it whenever you are summarizing something the JSON already cited.

---

## Template (fill literally)

```markdown
# RFP Reverse-Engineering Brief — <solicitation_id>

## 1. Inferred contract type & posture
- The CO published a <SOW|PWS> — <one-line signal> [see inferred_intake_answers.sow_or_pws.source_chunk_ids].
- Contract type is <FFP|T&M|CR|hybrid> — <one-line signal> [see inferred_intake_answers.contract_type.source_chunk_ids].
- Commercial status: <commercial|non-commercial> [see inferred_intake_answers.commercial_status.source_chunk_ids].
- **Our read:** <2-line judgment on what this posture rewards in our cost stack and bid strategy.>

## 2. Top hot buttons (max 5)
- HB-001 <label> [see hot_buttons[0].source_chunk_ids] — discriminator: <one-line play>.
- HB-002 <label> [see hot_buttons[1].source_chunk_ids] — discriminator: <one-line play>.
- HB-003 <label> [see hot_buttons[2].source_chunk_ids] — discriminator: <one-line play>.
- **In our capture experience,** <one-line pattern that ties HB-001..003 together.>

## 3. Top discriminator hooks (max 5)
- DH-001 on <evaluation_factor> — hook phrase "<phrase>" [see discriminator_hooks[0].source_chunk_ids] — proposal play: <one-line>.
- DH-002 on <evaluation_factor> — hook phrase "<phrase>" [see discriminator_hooks[1].source_chunk_ids] — proposal play: <one-line>.
- **Our read:** <one-line on which hook is the highest-leverage win-theme anchor.>

## 4. Top ghost-language risks (max 5)
- GL-001 "<phrase>" in <section> [see ghost_language[0].source_chunk_ids] — risk: <one-line>; Q&A action: <one-line>.
- GL-002 "<phrase>" in <section> [see ghost_language[1].source_chunk_ids] — risk: <one-line>; Q&A action: <one-line>.
- **Likely** the CO will <one-line on which gap they will close themselves vs leave open.>

## 5. CO errors detected
- FAR 52.237-2 trap: <status> [see key_personnel_clause_check] — recommended clause: <agency-appropriate alternative>.
- CPFF form ambiguity: <status> [see cpff_form_signal] — Q&A action: <one-line>.
- QASP↔CPARS confusion: <status> [see qasp_cpars_check] — recommended remedy: <one-line>.
- **Our read:** <one-line on what these errors signal about CO sophistication and amendment likelihood.>

## 6. Missing-section signals
- Section <N> absent — likely meaning: <one-line per the catalog inference table> [see missing_sections[0]].
- Section <N> absent — likely meaning: <one-line> [see missing_sections[1]].

## 7. Top clarification questions to file
- Q1: <one-line question> [see clarification_questions[0]].
- Q2: <one-line question> [see clarification_questions[1]].
- Q3: <one-line question> [see clarification_questions[2]].
- Q4: <one-line question> [see clarification_questions[3]].
- Q5: <one-line question> [see clarification_questions[4]].

## 8. Hand-off to proposal-generator
- Decision tree blocks 1–6 ready in `artifacts/rfp_reverse_engineering.json` [see decision_tree_reconstruction].
- Hot buttons + discriminator hooks ready to seed win themes [see hot_buttons, discriminator_hooks].
- Open clarifications flagged for the next Q&A round [see clarification_questions].
- **Our read:** <one-line on what proposal-generator should prioritize first when consuming this envelope.>
```

---

## Self-audit regex (apply mentally before returning)

Each line that begins with `- ` (bullet) MUST satisfy ONE of:

1. Ends with `]` (anchor tag — `[see …]`, `[chunk-…]`, `[entity: …]`, `[per references/…]`)
2. Begins with `- **Our read:**`, `- **In our capture experience,**`, `- **Likely`, `- **Classic`, `- **Rule of thumb:**`

If neither matches: rewrite to add an anchor (look up the path in the JSON envelope) OR demote to a Class B bullet with a marker OR delete the bullet. **Do not** invent chunk ids. **Do not** add a marker to a sentence that is actually a Class A fact (transparent and produces a worse brief).

Headings (`#`, `##`) and code fences are exempt.
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall