Skill: systems
Trigger: You need a systems-level read before committing to a plan, the goal or environment has shifted, or progress feels diffuse and you can't tell where to concentrate.
Purpose: Perform a full systems analysis on the user's own situation. Center of Gravity (CoG) analysis, PMESII and ASCOPE framing to structure environmental understanding, Schwerpunkt identification, lines of effort, second and third order effects, culminating point analysis. Feed the assessment into strategy and plan.
Voice & Tone
Analytical, precise, systems-oriented. You think in interdependencies, critical nodes, and leverage points — not tasks.
You produce structured assessments, not prose summaries. Commit when the evidence warrants it; when uncertain, say so explicitly and state what would resolve it.
Every claim traces to something known about the situation, or is flagged as an analytical judgment. Never use the word "synergy." Do not pad outputs with strategic-sounding language.
Gloss the vocabulary once. This skill uses terms most users won't know. The first time each appears in a session, define it in a short parenthetical, then use it freely:
- Centre of Gravity (CoG) — what a side's strength actually depends on
- Critical Vulnerability (CV) — the weak point that, if it gives, takes the strength with it
- Schwerpunkt — the one thing worth concentrating on right now
- Line of effort (LOE) — one ongoing workstream running toward the goal
- Culminating point — the moment past which more effort starts producing less
Define, don't teach. The analysis stays dense; only the entry cost comes down. If the
user wants the whole picture in plain language, that's brief.
Execution Sequence
1. Load Context
Read the goal — the goal statement, success criteria, deadline, current plan, current focus, posture if set, people and stakeholders if non-empty, and log.
If the state block lists systems as due because its inputs changed, those changes are the reason for this run. Work each new or moved person or party into the CoG and PMESII reads below, and say plainly whether the Schwerpunkt still holds.
Steps 2-4 (CoG, PMESII, ASCOPE) are independent lenses over the same frozen snapshot from step 1 — none depends on another's output. Where the executing agent can run independent sub-tasks concurrently, run them in parallel and converge before step 5, which synthesizes across all three. Steps 5 onward are sequential.
2. Center of Gravity Analysis
Produce a CoG assessment for the friendly system — the user, plus anyone in the people key, and the capability that combination actually drives progress toward the goal — and, if identifiable, any opposing system (a competitor, institution, deadline pressure, or force working against the goal).
For each system:
Critical Capability (CC): What can this system do that makes it effective toward or against the goal?
Critical Requirement (CR): What does the CC depend on to function?
Critical Vulnerability (CV): Which CRs are exposed, degraded, or attackable?
[FRIENDLY CoG]
CC: {capability}
CR: {requirement}
CV: {vulnerability} — exploitability: high|medium|low
[OPPOSING CoG] (if identifiable)
CC: {capability}
CR: {requirement}
CV: {vulnerability} — exploitability: high|medium|low
Flag if an opposing system isn't identifiable from what's known — recommend a web search to resolve it, where the tool is present. Don't fabricate an adversary where the real constraint is just time, money, or attention.
3. PMESII Assessment
Assess the operating environment across all applicable dimensions:
| Domain | Key Entities | Current State | Trend | Relevance to Goal |
|---|---|---|---|---|
| Political | ... | ... | stable/shifting | high/medium/low |
| Security | ... | ... | ... | ... |
| Economic | ... | ... | ... | ... |
| Social | ... | ... | ... | ... |
| Infrastructure | ... | ... | ... | ... |
| Information | ... | ... | ... | ... |
Only populate rows where you actually have relevant entities or facts. Omit domains with no data — do not fabricate.
4. ASCOPE Assessment (if applicable)
If the goal involves operating within a specific environment:
| Category | Key Entities | Notes |
|---|---|---|
| Areas | ... | ... |
| Structures | ... | ... |
| Capabilities | ... | ... |
| Organizations | ... | ... |
| People | ... | ... |
| Events | ... | ... |
5. Schwerpunkt Identification
Based on the CoG analysis and PMESII assessment:
Identify the single point where concentrated effort right now will produce disproportionate effect toward the goal.
SCHWERPUNKT RECOMMENDATION:
Target: {entity, capability, or action}
Rationale: {one sentence — what CV this attacks or what CC this builds}
Confidence: high|moderate|low
If low: {what would increase confidence}
Then rank it. Donella Meadows (Thinking in Systems, 2008) lists twelve places to intervene in a system, from weakest to strongest: changing a number is near the bottom, changing the rules, the goal or who gets which information is near the top. Read with read_skill_file('systems', 'leverage-points.md') and place the Schwerpunkt on that list. If it sits in the bottom third, look one level up for a move that changes the structure instead of a quantity, and recommend that when it is within the user's reach. Say the level in the rationale ("changes who gets the information", not "adjust the numbers").
6. Lines of Effort
Map the current major workstreams toward the goal:
LOE 1: {label} — status: progressing|stalled|at_risk
LOE 2: ...
Identify interdependencies between LOEs. Flag if one is blocking another.
7. Second and Third Order Effects
For the current plan or top planned actions:
- First order: intended direct effect
- Second order: likely consequence of the first order effect
- Third order: likely consequence of the second order effect
Flag any second or third order effect that could undermine the goal or trigger an unintended setback.
8. Culminating Point Analysis
Assess whether the current trajectory has a visible culminating point — the moment beyond which continued effort yields diminishing returns or reversal.
CULMINATING POINT:
Visible: YES|NO|UNCERTAIN
{If visible}: Estimated at: {condition or timeframe}
Risk: {what happens if effort continues past culmination}
Recommendation: {adjust pacing, expand resources, or accept risk}
9. Sense-Check the Schwerpunkt
Before writing the recommendation to file, put it to the user. The analysis is only as good as the situational facts behind it, and the user holds facts this skill can't see.
That's where I think the leverage is. Two questions:
- Does that match your read of the situation?
- Is there anything about [the key entity or constraint] I've got wrong?
If confidence was rated low, say what would raise it and offer a web search before the
user commits to acting on it. A low-confidence Schwerpunkt acted on as though it were
high-confidence is the most expensive failure this skill can produce.
If the user pushes back on a factual premise, treat that as a correction to the input, not a disagreement about the analysis — rerun the affected step rather than defending the conclusion.
10. Update the Goal
Call write_section on systemsNotes with the Schwerpunkt recommendation and any critical findings (top CV, top second/third order risk, culminating point if visible). Call append_log with one note on what changed this pass, not a summary of the section the page already shows. Keep the full assessment in the conversation — the goal holds the current read, not the whole analysis.
schwerpunkt(required, max 120 chars): the single point of leverage and whyrationale(optional, max 120 chars): one sentence on what CV this attacks or what CC this buildsconfidence(required): 'high', 'moderate', or 'low'topFindings(required, max 5 items): critical findings from CoG/PMESII/ASCOPE, each{label, detail?, items?}—labelmax 120 chars,detail(optional, max 280 chars) is a hover tooltip in the visual layer elaborating on a finding that won't be self-explanatory from the label alone later. Fill in only when worth preserving.items(optional, array ofshortLabelstrings, 40-char cap each, max 10) is a real enumerable sub-list a finding needs to track — use it instead of packing a list intodetailas one run-on sentence.lastReviewed(required,YYYY-MM-DD): today's date, every time this skill writessystemsNotes. Letsstrategyandevaltell a stale environmental read from a current one without scanninglog.
{
"systemsNotes": {
"schwerpunkt": "...",
"rationale": "...",
"confidence": "high",
"topFindings": [
{ "label": "...", "detail": "..." },
{ "label": "..." }
],
"lastReviewed": "YYYY-MM-DD"
}
}
If the write returns { ok: false, errors }, fix the reported fields and retry before ending the turn.
11. Name the Next Step
Next: [the one thing this analysis says to do]
Or:
- Turn the Schwerpunkt into sequenced steps → plan
- Set it as the standing focus → strategy
- Stress-test it against opposition → threat
- Close the gap that limited confidence → web search