policy-brief
Writes a policy brief: key message, evidence, risks, options and caveats in plain, honest language.
SKILL.md.grok/skills/policy-brief/SKILL.md
GitHub<!-- Generated from .claude/skills/policy-brief/SKILL.md by tools/sync_agents.py. Edit the source, then run the script. -->
/policy-brief — Policy Brief
Category: Documentation · Primary agent: policy-relevance-specialist
Writes a policy brief: key message, evidence, risks, options and caveats in plain, honest language.
Arguments: whatever the user wrote after the command name.
Purpose
Write a policy brief that turns model results into a decision-ready message: bottom line first, the evidence, the risks, the options and their trade-offs, with uncertainty and the most important caveats in the main text. Every number traces to a model run. The output is short, honest and readable by a senior decision-maker in minutes.
Use it when results must inform a decision or be shared with non-technical audiences.
Do not use it when a methodology document is needed (/technical-note).
Before you start
- Check required inputs. For each input below record
FOUNDorABSENT(never 'assumed'):- Model results
- Scenario/risk analysis
- Audience description
If a required input is
ABSENT, that part isNOT ASSESSED — NO DATA: do not estimate or invent it; name what is missing and which skill produces it. If all required inputs are absent, stop and say so.
- Check confidentiality. Identify the data classification level of the inputs (
.claude/docs/confidentiality-and-ai-use.md). Do not proceed with restricted or confidential data in a way the policy does not allow. - Collaboration. Follow
.claude/docs/collaboration-protocol.md: ask questions first, present options, get approval before writing files, and updateproduction/session-state/active.mdafter each approved section.
Agents
policy-relevance-specialist— primary agent; coordinates the work and integrates resultspolicy-brief-writer— Writes policy briefs and notesstakeholder-liaison-specialist— Interfaces with stakeholders and counterpartseconomic-reasoning-specialist— Frames the economic question and tests whether results make economic sense
Each agent is defined in .claude/agents/<name>.md. If your tool runs subagents, run these agents as subagents, in parallel where the steps are independent; otherwise read each agent's file and apply its instructions in turn. Subagents only write to the paths named in your instructions.
Key design decisions (present options, let the user choose)
| Decision | Typical options | What it changes |
|---|---|---|
| Audience | senior decision-makers, policy analysts, public | Tone and length |
| Length | one page, two to four pages, with annex | Detail |
| Uncertainty | ranges, scenarios, fan charts | Communication |
| Options | single recommendation, menu with trade-offs | Decision support |
Workflow
- Ask the audience, decision and length
- Draft bottom line, evidence, risks, options; keep technical detail in an annex
- Explain mechanisms in plain language without distorting them
- Check each number against the model run; include uncertainty ranges
- Obtain economic-reasoning-specialist and policy-relevance-specialist review; adjust to feedback
Checkpoints with the user
- After the bottom line: confirm the message
- Before release: confirm caveats and numbers
Common pitfalls
- Burying the message
- False precision
- Caveats in footnotes
- Mixing model results and judgment without labels
Output
docs/briefs/<name>-brief.md; a Word version or slides on request (python tools/report.py docx|pptx <file.md> --out <file>, charts embedded)
The brief has a bottom line, evidence, risks, options, uncertainty and caveats in the main text, and a technical annex; a number-tracing table supports review.
Gate / verdict
- Uncertainty and top-3 caveats are in the main text
- Every number traces to a model run
- Model results, assumptions and judgment are distinguished
Suggested next steps
/chart-pack