SITREP Style Guide for AI Agents
March 5, 2026 · View on GitHub
This style guide codifies the conventions and patterns observed across military SITREPs, intelligence situation reports, daily briefs, and open-source analytical products. It is intended as a reference for AI agents tasked with producing situational reports.
1. Core Principles
Clarity over elegance. Every sentence must convey actionable information. Avoid filler, hedging language, or rhetorical flourishes. The reader is a decision-maker under time pressure.
Objectivity and sourcing. Distinguish clearly between confirmed facts, assessed judgments, and unverified reporting. Always attribute claims to their source or basis.
Structured consistency. Use a fixed format for every report in a series. Readers rely on positional memory — they know where to find what they need.
2. Document Structure
2.1 Header Block
Every SITREP begins with a metadata header. Include all of:
- Report title/series name (e.g., "Israel SitRep," "Iran Update," "Maritime Intelligence Daily")
- Date and time of publication (use DD MMM YYYY format:
6 AUG 2001,2 MAR 2026) - Period covered if different from publication date (e.g., "Covering events of 25 Feb – 4 Mar 2026")
- Classification or sensitivity marking (even if unclassified, state it)
- Originating organization and author/editor if applicable
- Report number in a series (e.g., "043-26")
2.2 Bottom Line Up Front (BLUF) / Key Takeaways
Open with 3–7 numbered or bulleted key takeaways. These must:
- Be self-contained — each takeaway is understandable without reading the full report
- Lead with the most significant or time-sensitive development
- Use definitive language where the assessment is confident; qualify where it is not
- Avoid simply restating section headings — synthesize and assess
Example pattern:
- IDF strikes degraded approximately 60% of IRGC missile infrastructure in southeastern Iran.
- Iran shifted retaliatory strategy from ballistic missiles to drone swarms following launcher attrition.
- Strait of Hormuz traffic effectively ceased; only 5 commercial crossings recorded in 24 hours.
2.3 Body Sections
Organize by topic or theater, not chronology. Standard section types include:
- Situation Overview — Current state of the operating environment
- Operations — Friendly force actions, missions conducted, results
- Enemy/Adversary Activity — Opposing force actions, capabilities, intent
- Regional/Proxy Activity — Actions by allied or affiliated actors
- Logistics/Sustainment — Supply, equipment, infrastructure status
- Assessment/Outlook — Analyst judgment on trajectory and implications
Each section should open with a one-sentence summary before expanding into detail.
2.4 Closing / Commander's Assessment
End with a forward-looking assessment. This is where the analyst's judgment is most explicitly stated. Use expressions of likelihood (see Section 5).
3. Writing Conventions
3.1 Sentence Structure
- Lead with the subject and action: "IDF forces struck three IRGC compounds" not "Three IRGC compounds were struck by IDF forces"
- Standard pattern: Timeframe → Subject → Verb → Object → Source/Qualifier
- "On March 2, Hezbollah launched rockets toward Mishmar HaCarmel, according to IDF spokesperson statements."
- Keep sentences under 40 words
- One idea per sentence
3.2 Terminology
- Use precise military/technical terms; do not simplify for a general audience
- Spell out acronyms on first use, then use the acronym consistently
- Use proper names and designations: "IRGC Quds Force" not "Iranian special forces abroad"
- For weapons systems, include the designation: "Kornet ATGM (5,500m range, laser beam-riding)"
3.3 Numbers and Quantities
- Use numerals for all quantities: "3 compounds," "70 investigations," "5 crossings"
- Provide specific figures where available; use "approximately" only when the figure is genuinely estimated
- For casualty figures, distinguish confirmed from reported: "approximately 170 killed (per local media reports)"
3.4 Time References
- Use DD MMM YYYY format in headers:
6 AUG 2001 - In running text, use full dates on first mention, then abbreviated: "On March 2, 2026... Later on March 2..."
- Use 24-hour Zulu time for military reports:
1000Z,2000Z - When covering a period, state it explicitly: "During the reporting period (25 Feb – 4 Mar)..."
3.5 Geographic References
- Use the place name most recognizable to the target audience
- Provide geographic context on first mention: "Dahiyeh al-Janoubia (southern Beirut suburbs)"
- For maritime reports, include coordinates or sea area designations
4. Sourcing and Attribution
4.1 Source Descriptors
Attribute every non-obvious claim. Standard source descriptors:
- Official statements: "according to the IDF spokesperson," "per CENTCOM press release"
- Named media: "Al Jazeera reported," "as reported by Reuters"
- Social media / OSINT: "Telegram channels affiliated with [group] claimed," "satellite imagery reviewed by [org] shows"
- Intelligence sources: "clandestine reporting indicates," "a [REDACTED] service reported"
- Analyst judgment: "we assess," "ISW assesses," "this likely indicates"
4.2 Confidence and Reliability
When citing information of uncertain reliability:
- Qualify the source: "unconfirmed reports suggest," "a single-source report indicates"
- Do not present speculation as fact
- If multiple sources corroborate, say so: "multiple media outlets and Telegram channels confirm"
5. Expressions of Likelihood
Use a standardized probability scale when making assessments. The intelligence community standard (ICD 203):
| Expression | Probability |
|---|---|
| Almost No Chance | 1–5% |
| Very Unlikely | 5–20% |
| Unlikely | 20–45% |
| Roughly Even Chance | 45–55% |
| Likely | 55–80% |
| Very Likely | 80–95% |
| Almost Certain | 95–99% |
Use these terms precisely. "Likely" means 55–80% confidence, not "probably" in the colloquial sense.
6. Visual and Structural Formatting
6.1 Bullet Points
Use bullet points for:
- Lists of events, strikes, or actions
- Key takeaways and summaries
- Equipment catalogs and order-of-battle entries
Keep each bullet to 1–3 lines. Start each with an action or subject, not a gerund.
6.2 Bold and Emphasis
- Bold section headings and sub-headings
- Bold the names of key actors, operations, or events on first mention in a section
- Use bold sparingly in body text — only for the most critical facts a scanning reader must not miss
6.3 Tables
Use tables for structured comparisons:
- Weapons systems (type, range, generation, capability)
- Force dispositions (unit, location, status)
- Event timelines (date, event, source)
6.4 Maps and Graphics
Reference geographic or data visualizations when available. In text-only output, substitute with clear geographic descriptions and directional references.
7. Report Types and Their Conventions
7.1 Daily Situation Report (SITREP)
- Published on a fixed schedule (daily or twice-daily)
- Covers a defined reporting period
- Follows a rigid paragraph structure (see CJCSM 3150.05F for the 10-paragraph military format)
- Focuses on changes since the last report
7.2 Intelligence Summary (INTSUM / GRINTSUM)
- Synthesizes intelligence from multiple sources
- Includes analyst assessments with confidence levels
- May include graphical elements (maps, satellite imagery annotations)
- Article types: SIGACTs, Intelligence Highlights, Storyboards, Profiles, Historical Background Notes
7.3 Presidential Daily Brief (PDB)
- Extremely concise (typically 1–2 pages per topic)
- Written for a single consumer (head of state)
- Synthesizes the most critical intelligence across all domains
- No jargon — clarity for a non-specialist decision-maker
- Historical context woven into current reporting
7.4 Special / Crisis Report
- Published outside normal cycles in response to a significant event
- Opens with immediate situation, then expands to context and implications
- May combine elements of SITREP and INTSUM formats
7.5 Domain-Specific Reports (e.g., Maritime Intelligence Daily)
- Follow domain conventions (AIS data, shipping lanes, port activity)
- "At a Glance" summary followed by operational detail
- Quantitative data emphasized (vessel counts, traffic volumes, disruption metrics)
8. Common Pitfalls
- Burying the lead. The most important development goes first, not last.
- Passive voice. "Strikes were conducted" obscures who acted. Name the actor.
- Vague attribution. "Reports indicate" — which reports? From whom?
- Overqualification. If the evidence is strong, state the assessment directly. Reserve hedging for genuinely uncertain judgments.
- Chronological narration. Organize by significance and topic, not by the order events occurred.
- Mixing fact and assessment. Keep reporting (what happened) and analysis (what it means) clearly separated or explicitly labeled.
- Inconsistent formatting. Every report in a series must use the same structure. Readers depend on it.
9. Checklist for AI Agents
Before finalizing a SITREP, verify:
- Header block is complete (title, date, period, classification, source, number)
- BLUF/Key Takeaways section leads the document with 3–7 self-contained points
- Every factual claim is attributed to a source
- Assessments use standardized likelihood language
- Acronyms are defined on first use
- Sentences average under 40 words
- Active voice is used throughout
- Body sections are organized by topic, not chronology
- Numbers are specific where data exists, qualified where estimated
- Time references use consistent formatting
- The report ends with a forward-looking assessment
- Format matches the series template exactly