agentleFS
tatashbartatashbar/Economic-Modeling-Studio/.agents/skills/build-fpas/SKILL.md

build-fpas

Skill4 starsChanged 6 days agoMITView on GitHub

Designs and builds a forecasting and policy analysis system: core model, nowcasts, judgment layer, forecast process.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do not use it when: only a model is needed (/build-semi-structural, /build-dsge) or a single forecast is needed (/forecast-round).

    line 20
  2. Very specific:

    If a required input is ABSENT, that part is NOT ASSESSED — NO DATA: do not estimate or invent it.

    line 26
  3. Very specific:

    Collaboration: Follow .claude/docs/collaboration-protocol.md: ask questions first, present options, get approval before writing files, and …

    line 28
  4. Distinctive:

    Run a mock forecast round on historical data and review lessons

    line 54
  5. Distinctive:

    Use a frozen vintage; evaluate the mock forecast against outturns

    line 55
  6. Distinctive:

    After the process design: confirm calendar and roles with the institution

    line 61
  7. Distinctive:

    After integration: confirm the overlay rules

    line 62
  8. Distinctive:

    After the mock round: agree changes before handover

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md
<!-- Generated from .claude/skills/build-fpas/SKILL.md by tools/sync_agents.py. Edit the source, then run the script. -->

/build-fpas — Build FPAS

Category: Build · Primary agent: fpas-modeler

Designs and builds a forecasting and policy analysis system: core model, nowcasts, judgment layer, forecast process.

Arguments: whatever the user wrote after the command name.

Purpose

Design and build a forecasting and policy analysis system: not only a core model but the process around it, including nowcasting tools for the near term, a judgment layer with an overlay log, forecast-story templates, a calendar with roles and sign-offs, and the training that lets a counterpart team run it. The output is a system that completes a full forecast round from a frozen data vintage.

Use it when an institution needs a regular, documented forecast process tied to policy decisions, or an existing process needs a model core, a judgment layer or better documentation.

Do not use it when only a model is needed (/build-semi-structural, /build-dsge) or a single forecast is needed (/forecast-round).

Before you start

  1. Check required inputs. For each input below record FOUND or ABSENT (never 'assumed'):
    • Core model or approved model-spec
    • Data pipeline If a required input is ABSENT, that part is NOT 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.
  2. 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.
  3. Collaboration. Follow .claude/docs/collaboration-protocol.md: ask questions first, present options, get approval before writing files, and update production/session-state/active.md after each approved section.

Agents

  • fpas-modeler — primary agent; coordinates the work and integrates results
  • dfm-nowcasting-specialist — Builds dynamic factor models and nowcasts
  • stakeholder-liaison-specialist — Interfaces with stakeholders and counterparts
  • visualization-dashboard-specialist — Designs charts and dashboards

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)

DecisionTypical optionsWhat it changes
Core modelsemi-structural gap model, small DSGE, suiteTransparency versus structure
Near-term toolsbridge equations, DFM nowcasts, expert judgmentAccuracy of initial conditions
Judgment layeradd-factors, tuned shocks, conditioning on pathsTraceability of overlays
Cyclemonthly, quarterly; number of meetings per roundWorkload and timeliness
Outputsforecast report, risk scenarios, fan charts, policy rate pathCommunication

Workflow

  1. Define the forecast process calendar, roles and sign-offs
    • Data cut-off, nowcast, model runs, judgment meetings, sign-off, publication
  2. Integrate core model, near-term nowcast tools and judgment-overlay log
    • Overlays recorded with reason, evidence and owner
  3. Build the shock-decomposition and forecast-story templates
    • Templates link every story element to a decomposition or overlay
  4. Run a mock forecast round on historical data and review lessons
    • Use a frozen vintage; evaluate the mock forecast against outturns
  5. Prepare training material for the counterpart team
    • Hands-on exercises on the core model and the process (/design-training)
  6. Register the model as a lego set (/lego-map): a baseplate plus bricks with declared studs (variables, units, timing) and stubs, so components can later be attached, detached or swapped

Checkpoints with the user

  • After the process design: confirm calendar and roles with the institution
  • After integration: confirm the overlay rules
  • After the mock round: agree changes before handover

Common pitfalls

  • A model without a process, or a process without documentation
  • Overlays that are not recorded
  • Near-term tools inconsistent with the core model's initial conditions
  • Training that ends at the workshop with no follow-up

Output

models/fpas/<name>/ and docs/models/<name>/fpas-design.md (+ model.lego.json)

The model folder holds the core model, nowcast tools, templates and run scripts; the design document covers the calendar, roles, overlay rules and the mock-round review; the lego manifest registers the model and its satellites. Every run saves its results to output/<run-id>/ in the model folder, with a run log (.claude/rules/model-code.md).

Gate / verdict

  • A full mock round runs end-to-end from a frozen data vintage
  • Lego manifest passes python tools/lego.py check
  • The overlay log is in place and used in the mock round
  • Training material exists for the counterpart team

Suggested next steps

  • /forecast-round
  • /design-training
  • /lego-map

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.