Back to profile
Case Study: Agenda-Intelligence.md
TL;DR
- Built Agenda Intelligence MD as a machine-readable Markdown protocol and CLI toolkit that turns news scanning into decision-ready agenda analysis.
- Developed a Codex-ready workflow for source-backed briefs across policy, sanctions, regulation, geopolitics, markets, and strategic risk.
- Added source plans, JSON schemas, evaluation assets, AnalysisBank memory patterns, and regional lenses for Central Asia + Caspian and Middle East.
- Optimized the repo for policy analysts, sanctions/compliance teams, market-risk analysts, founders, and AI agents that need useful analysis rather than generic summaries.
Evidence
- Public GitHub repository: github.com/vassiliylakhonin/agenda-intelligence-md
- Agent-readable orientation file:
llms.txt.
- Core protocol:
Agenda-Intelligence.md.
- CLI package:
agenda-intelligence-md.
- Schemas: brief, evidence pack, and manifest validation.
- Source plans: sanctions, regulation, elections, conflict, technology/AI, markets, and regional monitoring.
- Evaluation assets: rubric, LLM judge prompt, human checklist, and sample cases.
Project state (self-reported)
- Latest release: v0.7.1 (see GitHub releases for current).
- Distribution: PyPI package (
agenda-intelligence-md), GitHub release wheel, editable source install, and plain Markdown protocol attachable to any agent.
- Implemented layers: Markdown protocol, source acquisition plans, JSON schemas (brief / evidence pack / manifest), CLI validation (
validate, doctor, guided start), MCP server, evaluation toolkit (rubric, LLM-judge prompt, human checklist, before/after sample cases).
- Focus domains: policy, sanctions, regulation, geopolitics, markets, conflict, elections, and strategic risk.
- No production-usage, adoption, or benchmark numbers are claimed.
Context/Constraint
AI agents are getting better at collecting information, but many still write weak news analysis.
The common failure mode is simple: they recap events, add confident-sounding commentary, and stop before the answer becomes useful.
The project needed to work as a small reusable layer, not as a long prompt that bloats every agent context.
Problem
Most AI-generated public-agenda analysis is readable but decision-light. It rarely says what changed, who gained leverage, what remains unknown, or what would falsify the view.
That is fine for a summary. It is weak for research, compliance, policy monitoring, investment context, or operating decisions.
Actions
- Reframed the project as a Markdown protocol and CLI toolkit for agenda analysis, not a news-summary skill.
- Defined the core separation agents should preserve:
Fact → Assessment → Assumption → Unknown → Scenario → Indicator to watch
- Added triage categories: noise, weak signal, signal, structural shift, trigger event, compliance-relevant development.
- Added evidence rules so agents do not imply live verification when none happened.
- Added output patterns for compact briefs, decision memos, red-team checks, and watchlists.
- Added CLI onboarding with a
start command that prints a trimmed source plan, brief template, and next validation commands.
- Added JSON schema validation so briefs, evidence packs, and manifests can be checked before use.
- Added evaluation assets so outputs can be scored against clarity, evidence discipline, uncertainty, and decision usefulness.
- Added lightweight regional lenses (Central Asia + Caspian, Middle East) at the protocol level — disambiguated from the standalone Central Asia + Caspian Hybrid Intelligence Skill, which carries the deep specialist reasoning.
- Kept the adoption path Codex-friendly while preserving a runtime-agnostic protocol layer.
What it does now
- Helps agents avoid shallow news summaries.
- Forces a clear answer to “what changed?”
- Separates facts from judgments and assumptions.
- Makes uncertainty visible without turning the answer into vague caveats.
- Produces watch-next indicators instead of soft endings.
- Gives regional checklists for higher-context analysis.
- Gives Codex and other agents a stricter workflow for source-backed monitoring and policy-risk memo generation.
Current outputs
- Bottom line.
- Signal classification.
- What changed.
- Why it matters.
- Who is affected.
- Main uncertainty.
- Scenarios.
- Watch-next indicators.
Regional lens packs
Central Asia + Caspian
Focused on sanctions routing, corridor politics, Caspian chokepoints, banking/payment exposure, state leverage, energy, minerals, and regional political economy.
Middle East
Focused on escalation risk, energy flows, maritime chokepoints, sovereign capital, sanctions exposure, normalization, and regional power competition.
Who it is for
- People building research or news agents.
- Policy and geopolitical analysts.
- Compliance and sanctions-monitoring workflows.
- Investors and founders tracking operating-context shifts.
- NGOs and advisory teams monitoring regional risk.
- Anyone tired of agents producing fluent but low-value news commentary.
Why this version is better
It is small enough to reuse and strict enough to change the output. The protocol does not ask an agent to sound smarter. It asks the agent to classify signal, expose uncertainty, and name what to watch next.
That is the part most generic news analysis misses.
Tech stack
- Plain markdown.
- Python CLI package.
- JSON schemas.
- Evaluation toolkit.
- GitHub repository with CI validation.
- Agent-readable
llms.txt.
- Codex-ready integration path.
Relevance
This project demonstrates how I think about useful agent infrastructure: small files, explicit reasoning contracts, low context cost, and outputs that improve decisions rather than just sounding polished.
Before / after (illustrative)
Evidence mode: reasoning-only. Excerpt from a worked example in the repo, condensed for this page. Full example with diagnosis and side-by-side table: examples/before-after/sanctions-routing.md.
User question: "Reports say more restricted goods may be moving through Central Asia to Russia. What does this mean?"
Before — generic agent output:
Reports of restricted goods moving through Central Asia highlight the importance of sanctions compliance and supply-chain monitoring. Companies should ensure they conduct due diligence and monitor regulatory developments.
Fluent and correct-sounding. Contains no signal classification, no specific transmission mechanism, no main uncertainty, and no falsifiable indicators.
After — with the Agenda Intelligence MD protocol attached:
- Bottom line: Treat as a compliance-relevant signal unless enforcement action, customs data, named firms, or regulator guidance escalate it. Key question: does suspected routing become a documented enforcement pattern.
- Signal classification: Weak signal → signal if supported by customs anomalies, designations, bank behavior, or named intermediaries.
- What changed: Central Asia is not a passive route; the relevant mechanisms are traders, customs brokers, banks, logistics firms, beneficial-ownership structures, and state tolerance for re-export revenue versus sanctions pressure.
- Main uncertainty: Whether the reports reflect isolated evasion cases, statistical noise, or a repeatable routing system.
- Watch next: sanctions designations; customs data by HS code; bank account closures; export-control advisories; named intermediaries; correspondent-banking behavior; customs-inspection changes.
The protocol does not make the agent sound smarter. It forces it to classify the signal, expose uncertainty, and name what would update the view.
Portfolio context
Agenda Intelligence MD is the infrastructure / validation layer in a three-repo portfolio:
- Infrastructure / validation — Agenda Intelligence MD (this case study). Schemas, validation, scoring, evidence audit, CLI / MCP / CI tooling.
- Horizontal domain skill — Global Think Tank Analyst. Reasoning method and memo modes, region- and topic-agnostic.
- Vertical specialist — Central Asia + Caspian Hybrid Intelligence Skill. Region-deep reasoning that rides on top of the horizontal skill and validates through this repo.
This repo does not duplicate either neighbor. Reasoning method and regional depth live in the skill repos; this one provides the contracts, validators, and tooling.
Project links
Author: Vassiliy Lakhonin