Policy Brief Drafter

intent

use this skill when a government-affairs analyst, think-tank researcher, or advocacy staffer needs to convert legislative text, regulatory dockets, studies, hearing transcripts, expert interviews, and stakeholder notes into a 2, 4 page brief for a specific decision-maker on a specific decision with a specific deadline. the brief lands the ask (vote yes, file a comment, adopt a position, authorize funding) by laying out the problem, options, recommendation, counter-positions, and evidence in a format that decision-makers can act on in 5 minutes. use this skill only when you have a named audience, a concrete ask, and a calendar date or named milestone. do not use this skill to draft op-eds, fact sheets, general position papers, or legal opinions.

inputs

required before drafting:

1. audience (specific decision-maker) , a named committee chair and staff, a named regulator at a specific agency, the CEO and board of a member organization, the executive principal at a cabinet agency. generic "policymakers" or "the public" do not count.

2. ask (the exact decision) , vote yes on HR-1234 as amended, file a comment supporting Option B in docket X, adopt the coalition position, authorize $X for pilot, table the rulemaking, request a hearing. "raise awareness" is not an ask.

3. decision window , vote/hearing/comment-deadline date, or a named milestone tied to a calendar. "next cycle" is not acceptable; push for a specific date.

4. brief type , Informational (no ask, facts only), Advocacy (single recommendation), or Comparative options memo (analyze 3-5 options, recommend among them).

5. disclosure posture , Internal-only, Member-distribution, or Public-record. controls what strategy language and coalition details appear in the final brief.

6. length cap , 2 pages, 3 pages (default), or 4 pages, excluding appendices.

7. authority anchor , the statute, regulation, court decision, or executive order on which the decision rests. required to ground the brief in legal or regulatory reality.

8. evidence batches , user provides sources in batches, each tagged with title, author/issuing body, year, source class (Primary / Peer-reviewed / Gray literature / Advocacy / News / Expert interview), attribution level (if interview), and URL or full citation. you do not fabricate sources.

9. counter-evidence , at least one published source representing the strongest opposition to the ask. required for Advocacy and Comparative briefs; absence downgrades the brief to Informational.

10. audience evaluation criteria , the named decision-maker's priorities (effectiveness, cost, equity, legal authority, feasibility, political viability) against which options are scored. defaults apply if not stated.

procedure

phase 1: audience and ask (steps 1, 3)

step 1: lock the three anchors

ask the user, one question at a time:

1. "who is the specific decision-maker, and what is their title and organization?"
2. "what is the exact decision you need them to make, and when must they make it (calendar date or named milestone)?"
3. "what is the legal or regulatory anchor (statute, regulation, court case, executive order) for this decision?"

do not proceed until all three are confirmed. if the user offers "raise awareness" or "inform the general public," stop and reframe: "a policy brief targets a specific decision-maker on a specific decision. would you like to draft an op-ed or fact sheet instead?" (note: op-eds and fact sheets are out of scope for this skill.)

step 2: confirm brief type and disclosure posture

ask:

1. "is this brief Informational (facts only), Advocacy (single recommendation), or Comparative (analyze options and recommend)?"
2. "what is the disclosure posture: Internal-only (not shared outside the organization), Member-distribution (shared with coalition/membership), or Public-record (may appear in filed comments or public testimony)?"
3. "length cap: 2 pages, 3 pages, or 4 pages?"
4. "what are the decision-maker's main evaluation criteria? or should i use the defaults: effectiveness, cost, equity, legal authority, feasibility, political viability?"

step 3: restate and confirm

echo back, word-for-word: audience (name and title), ask (the decision being requested), decision window (date or milestone), brief type, disclosure posture, length cap, authority anchor, and evaluation criteria. get explicit user confirmation before moving to Phase 2. if the user says "close enough" or "mostly right," ask them to specify what you got wrong.

---

phase 2: evidence intake and source discipline (steps 4, 7)

step 4: ingest in tagged batches

say: "provide evidence in batches. for each item, give me the title, author or issuing body, year, source class, and a URL or full citation. i will track every quantitative claim and direct quotation back to one of these items. no exceptions."

for each source, capture:

field	required
title	yes
author / issuing body	yes
year	yes
source class	yes: Primary (statute, regulation, court decision, hearing transcript, official data series), Peer-reviewed, Gray literature (working paper, official report, think-tank brief), Advocacy / industry, News / opinion, Expert interview
attribution (if interview)	Attributed (on the record), On background, Off the record
URL or full citation	yes , user supplies; never fabricate

assign each source a stable ID (e.g., CBO-2024-01, HT-2023-05, EXP-JONES-2024) and add it to a tracking table as intake proceeds.

flag edge cases and gaps: "i see you have CBO fiscal data from 2024. do you have more recent agency enforcement data, or should i note this as a 2024 snapshot?"

step 5: counter-evidence and dissent sweep

before drafting, ask: "what are the strongest published counter-positions to the ask? provide at least one source for each, with author, year, and URL."

if the user cannot or will not supply counter-evidence, say: "a one-sided advocacy brief without acknowledged opposition is not credible. i am downgrading this to Informational type (facts only, no ask) until you provide at least one counter-source. alternatively, would you like to add counter-evidence?" do not proceed with Advocacy or Comparative drafting without counter-evidence.

step 6: source quality flags

classify each source:

• strong , primary (statute, regulation, official data, court decision) or peer-reviewed
• acceptable , gray literature (official report, think-tank brief, working paper) or attributed expert interview
• weak / contested , advocacy/industry publication, on-background interview, single news outlet, opinion without underlying source
• reject , blog/opinion without source, anonymous social media, unverifiable claim

rule: weak sources may appear in the brief only if corroborated by at least one Strong or Acceptable source, or explicitly labeled in-text as "industry estimate" / "advocacy claim" / "one news report".

flag any Weak or Reject sources before drafting: "source X is a single opinion piece from a partisan outlet with no underlying data. should i drop it, find a corroborating source, or label it as advocacy?"

step 7: quantitative claim routing

for every number that will appear in the brief, build a row:

claim	source ID	source class	status
"$X billion annual cost"	CBO-2024-01	Primary	sourced
"Y% of facilities non-compliant"	pending	pending	pending source

anything marked Pending source at draft time is dropped from the body and moved to "open items" at the end. never paraphrase an unsourced claim into the brief.

---

phase 3: options analysis (steps 8, 9)

step 8: build the options table

include 3, 5 options. always include Status quo as Option 0. if Advocacy or Comparative type, include the Recommended option.

score each option against the audience's named evaluation criteria (or use defaults: effectiveness, cost, equity, legal authority, feasibility, political viability).

option	effectiveness	cost	equity	legal authority	feasibility	political viability
0. status quo	[source]	[source]	[source]	[source]	[source]	[source]
1. [option name]	[source]	[source]	[source]	[source]	[source]	[source]
2. [option name]	[source]	[source]	[source]	[source]	[source]	[source]

each cell is 1-2 short sentences with a source tag , e.g., [CBO 2024, p.14] or [agency docket comment, Smith 2024]. any cell without a source must be flagged and resolved before drafting.

after the table, write 1-2 paragraphs per option explaining the trade-offs and implementation pathway.

step 9: name the recommendation and its falsifiers

for Advocacy and Comparative briefs only:

• state the recommended option in one sentence.
• state the falsifying conditions that would change it. example: "recommendation would shift to Option 2 if the cost estimate exceeds $X billion, or if the statutory authority is struck down by the D.C. Circuit."

do not bury the ask. the first 100 words of the executive summary must contain the decision-maker, the decision being requested, and the decision-window date.

---

phase 4: brief draft and talking points (steps 10, 12)

step 10: draft to the length cap

honor the length cap from step 2. follow this structure:

1. executive summary (≤150 words) , one sentence on the problem, one sentence on the ask with decision-window date, one sentence on the recommendation. write the first 100 words as a hook for the decision-maker: who are you, what do you need to decide, and by when.

2. problem statement , what is broken, for whom, and the magnitude. use cited numbers. avoid jargon.

3. background , the authority anchor (statute, regulation, case), prior actions, the current trigger event. answer: why are we deciding this now?

4. options analysis , the table from step 8 plus 1-2 paragraphs per option.

5. recommendation , the recommended option, why it scores best on the audience's criteria, and the falsifying conditions from step 9.

6. risks and trade-offs , the strongest counter-positions from step 5, named by author/organization and engaged directly. example: "Smith (2024) argues that Option 1 imposes undue compliance burden on small firms. however, [response citing source]. the risk is [source]."

7. evidence and source map , a table of every source cited, with citation, source class, and the section(s) where it appears.

8. decision window , the date, milestone, prerequisite actions, and what happens after the decision-maker acts.

step 11: talking-points and Q&A appendix

add as a separate section or page:

• talking points , ≤10 single-sentence statements covering: the ask, the top three reasons to act, the strongest counter and the response to it, the deadline.
• likely questions , 5-8 questions the audience will ask (e.g., "what if we don't have budget?", "what's the litigation risk?", "who else supports this?"). answer each in 1 paragraph with a source tag.

step 12: self-check gate

before output, verify all items below. if any check fails, return to the relevant step and ask the user for clarification.

• audience (name and title), ask (decision being requested), and decision-window date are present in the first 100 words of the executive summary.
• every quantitative claim in the body text is cited with a source tag. no Pending-source claims remain in the body.
• counter-positions from step 5 are acknowledged and engaged in the Risks and Trade-offs section. if they are missing, the brief is Informational type, not Advocacy.
• no fabricated citations, URLs, quotations, or anonymous sources. every source was provided by the user.
• disclosure posture is honored. internal-only strategy language does not appear in public-record or member-distribution versions. coalition preferences and member dissent are internal-only.
• length is within the page cap from step 2, excluding appendices.
• the brief does not offer legal advice, predict vote counts as fact, make partisan attacks on named individuals, or guarantee legislative outcomes. use "likely / contested / possible" framings with sources.
• the output is marked DRAFT , for policy lead review and awaits sign-off before final distribution.

---

decision points

decision point 1: is the audience a named decision-maker with a specific decision and a calendar date?

• if yes: proceed to step 2.
• if no (e.g., "raise awareness" or "inform the general public"): stop. say: "a policy brief targets a specific decision-maker on a specific decision. this skill does not cover op-eds, fact sheets, or general position papers. would you like to reframe your request, or is there a named decision-maker and decision-window you can point to?" do not draft.

decision point 2: does the user supply counter-evidence to the ask?

• if yes, Advocacy or Comparative type: proceed to step 6 with Advocacy or Comparative type intact.
• if no, Advocacy or Comparative type requested: downgrade to Informational type (facts only, no recommendation). say: "a one-sided advocacy brief without acknowledged opposition will not survive scrutiny. i am shifting this to Informational type until you provide at least one counter-source. alternatives: (1) add counter-evidence; (2) accept Informational type; (3) reframe the ask as a question ('should we adopt Option A?') rather than a directive."

decision point 3: does every quantitative claim have a sourced, non-Pending status before drafting?

• if yes: proceed to step 10 (draft).
• if no: flag the Pending claims. say: "these claims are still waiting for sources: [list]. should i (1) drop them from the body and add to 'open items'; (2) wait for you to provide sources; or (3) rephrase as a question ('preliminary analysis suggests $X, pending final data')?". do not draft until resolved.

decision point 4: is the brief Advocacy or Comparative type without falsifying conditions named?

• if yes: step 9 requires you to state the falsifying conditions (e.g., "recommendation flips to Option 2 if the cost exceeds $Y"). do not draft without them.
• if no (Informational type): proceed to step 10.

decision point 5: is the disclosure posture Public-record or Member-distribution, with internal-only strategy language in the draft?

• if yes: strip internal-only language (coalition preferences, member dissent, negotiation strategy, confidential polling). rewrite as neutral, fact-based analysis.
• if no: proceed to output.

decision point 6: does the brief contain legal advice, vote-count predictions stated as fact, or partisan attacks on named individuals?

• if yes: redact or rephrase. example: instead of "this will fail" say "this faces opposition from [named group] citing [source]". instead of legal interpretation, say "the statute is contested; [Group A] argues [interpretation with source]; [Group B] argues [interpretation with source]."
• if no: proceed to output.

---

output contract

format: markdown with YAML frontmatter.

structure (mandatory sections):

# Policy Brief ,  DRAFT (for policy lead review)

**Audience:** [decision-maker name and title]
**Ask:** [the decision being requested]
**Decision Window:** [date or milestone]
**Brief Type:** Informational / Advocacy / Comparative
**Disclosure Posture:** Internal-only / Member-distribution / Public-record
**Length:** [pages, excluding appendices]
**Authority Anchor:** [statute, regulation, case, or executive order]

---

## Executive Summary
[≤150 words; first 100 words contain audience, ask, decision-window date]

## Problem Statement
[Magnitude and impact, cited numbers]

## Background
[Authority anchor, prior actions, trigger event]

## Options Analysis

| Option | Effectiveness | Cost | Equity | Legal Authority | Feasibility | Political Viability |
| --- | --- | --- | --- | --- | --- | --- |
| 0. Status quo | [source] | [source] | [source] | [source] | [source] | [source] |
| 1. [Option name] | [source] | [source] | [source] | [source] | [source] | [source] |
| 2. [Option name] | [source] | [source] | [source] | [source] | [source] | [source] |

[1-2 paragraphs per option]

## Recommendation
[Recommended option + falsifying conditions]

## Risks and Trade-offs
[Strongest counter-positions named and engaged with sources]

## Evidence and Source Map

| Source ID | Citation | Source Class | Used in Section |
| --- | --- | --- | --- |
| [ID] | [Full citation] | Primary / Peer-reviewed / Gray / Advocacy / News / Interview | [Executive Summary / Problem / Options / etc.] |

## Decision Window
[Date, milestone, prerequisite actions, what happens next]

---

# Talking Points Appendix

## Talking Points (≤10 single sentences)

1. [ask + decision-window date]
2. [reason 1]
3. [reason 2]
4. [reason 3]
5. [strongest counter + response]
6. [deadline]
7-10. [additional context]

## Likely Questions

**Q. [Question 1]**
**A.** [1-paragraph answer with source tag]

**Q. [Question 2]**
**A.** [1-paragraph answer with source tag]

---

## Open Items
- [Pending-source claims dropped from body]
- [Counter-evidence still needed]
- [Authority questions for legal review]

data format:

• executable markdown with linked sources (URLs in citations).