systemic-protagonist

name

systemic-protagonist

description

Renders a market, protocol, institution, codebase, supply chain, disease, or ecosystem as a legible protagonist without faking agency. Builds the POSIWID character sheet (revealed objective function vs. stated purpose), the agency ledger of verb repairs, the gateway-proxy admission test, the traveling-object device, the instrumented threshold scene, and climax by constraint violation. Use when the subject of a piece is a system rather than a person, when a draft says "the market decided" or "the industry learned", when analysis reads as a data dump with no protagonist, or when user mentions systemic narrative, system as character, POSIWID, anthropomorphism, agency laundering, follow the flow, or institution as protagonist.

Systemic Protagonist

Table of Contents

• Core Principles
• Workflow
• The POSIWID Character Sheet
• The Agency Ledger
• Climax Types for a Subject With No Battle
• Domain Failure Profiles
• Guardrails
• Quick Reference

Related skills: Use narrative-evidence-ledger for typed claims before you start, narrative-form-triage for the story/explanatory/gathering decision, narrative-fallacy-guard for the inevitability and survivorship audit, mcphee-structure-derivation for arrangement when no dramatic engine exists, narrative-opposition-web for the opposition slot. This skill does not build that slot. In a system narrative corner B is a constraint, and its value cluster arrives by delegation to a documented human carrier.

Core Principles

1. A system is a character because it has a revealed objective function: not because it has feelings. Constraints, feedback loops, and a consistent conversion of inputs into outputs are what make behaviour predictable, and predictable behaviour is what makes a subject dramatizable.
2. Purpose is deduced from behaviour, not rhetoric: what a market pays for, what an institution promotes people for, what a codebase makes easy. The stated purpose is evidence about the speaker, not about the system.
3. The gap is the engine: stated purpose minus revealed objective function. No gap means no narrative — it means a description, and you should say so rather than manufacture tension.
4. Agency is a grammatical property before it is a narrative one: subject-verb-object smuggles intent. "The market wanted higher yields" is not a style problem, it is a false causal claim wearing a sentence.
5. Relocate agency, never delete it: the repair is naming who or what actually acted. Agentless passive ("prices were observed to decline") obscures causation worse than the anthropomorphism it replaced.
6. Systemic framing must be additive to accountability, not a substitute: "the system failed" is often a sentence written to avoid naming someone who had the authority and the information to do otherwise.
7. Humans in a systems piece are probes, not protagonists: their arcs exist to map the shape of the constraint they collide with. When the probe eats the system, you have a character study wearing a systems title.

Workflow

Copy this checklist and track your progress:

Systemic Protagonist Progress:
- [ ] Step 1: Gate — confirm subject level, evidence depth, and domain failure profile
- [ ] Step 2: Build the POSIWID character sheet (two regimes, mandatory residual)
- [ ] Step 3: Admit the humans and the traveling object
- [ ] Step 4: Stage perceptibility — threshold scenes and the constraint-violation peak
- [ ] Step 5: Write the billboard; run agency ledger and responsibility check
- [ ] Step 6: Drift check and handoff

Step 1: Gate

Step 1.1: Declare the subject level in one line: is the protagonist the SYSTEM, or a named participant inside it? Do not oscillate. If it is a participant, stop — this is the wrong skill.

Step 1.2: Run the material-depth gate. Threshold scenes, sense of place, and gateway arcs all require artefacts: timestamps, readings, transcripts, memos, logs, filings. Count them. If the corpus is entirely secondary and abstract, take the DOWNGRADE PATH: produce the character sheet and the billboard only, mark scene work as INSUFFICIENT GROUND-LEVEL MATERIAL, and recommend explanatory exposition. Never generate sensory detail that is not in the source.

Step 1.3: Select the domain failure profile from Domain Failure Profiles and name the dominant local error. One uniform checklist misses it.

Step 2: Build the POSIWID character sheet

Step 2.1: Answer all eight questions in The POSIWID Character Sheet, citing evidence per line.

Step 2.2: Apply the TWO-REGIME TEST. The character sentence must be supported by behaviour in at least two distinct regimes — boom and bust, peacetime and stress, normal operation and incident, funded and unfunded. Derived from one episode, the "revealed purpose" is a description of that episode dressed as a trait. One regime = UNSUPPORTED, go back to the material or widen the time window.

Step 2.3: Fill the residual field: which outcomes does the revealed objective function FAIL to explain? At least one entry is mandatory and it must survive into the finished piece. A character sentence that explains 100% of outcomes with zero residual is fitted noise, and it converts a system that was merely incompetent, path-dependent, or unlucky into a coherent malign optimiser.

Step 2.4: Name the gap in one sentence. If there is no gap, emit NO DRAMATIC ENGINE with the reason and recommend a different subject or a longer window. This is a successful outcome, not a failure.

See resources/posiwid-character-sheet.md for the field-by-field procedure, three filled examples, and the reform-attempt ledger.

Step 3: Admit the humans and the traveling object

Step 3.1: Run the gateway-proxy admission test on every candidate human. Each gateway must map to ONE line of the character sheet, must have started wrong and updated, and must carry a written paragraph on what they got wrong or could not see.

Step 3.2: Run the DELETION TEST and the RATIO TEST. Delete the gateway: does the systemic claim still stand on cited evidence? Count paragraphs of gateway interiority against paragraphs of mechanism.

Step 3.3: Disclose the cast selection in the text. State how each person came to be in the story and who declined, refused, or was unreachable — and whether their absence flatters anyone. This is non-optional.

Step 3.4: Choose the traveling object: the smallest concrete thing that moves end to end. Verify its four constraints and run the GIMMICK TEST.

See resources/entry-points-and-peaks.md for both tests and the station-list procedure.

Step 4: Stage perceptibility

Step 4.1: Build the instrumented threshold scene at instrument resolution, constrained to what the observer could know at that timestamp, evidenced by a document created at or before it.

Step 4.2: State whether the threshold is a property of the SYSTEM or of the MEASUREMENT.

Step 4.3: Select the peak from the five legitimate climax types and run the TRANSFER-FUNCTION TEST: after this, do the same inputs produce different outputs? If no, it is a scene, not a climax.

Step 4.4: Extract every REFORM ATTEMPT and include at least one that partially worked, or an explicit statement of the conditions under which this system has changed. Without it the piece becomes fatalism, which is inevitability with the sign flipped.

Step 5: Billboard, verbs, and responsibility

Step 5.1: Write the five-sentence stated-vs-revealed billboard. State the finding; do not tease it.

Step 5.2: Emit the agency ledger as a binding verb-constraint sheet for the drafting stage. If a draft exists, run the repairs over it and then run the POST-PASS CHECK on agentless passives.

Step 5.3: Run the responsibility-laundering reverse check on every systemic attribution: did a named actor have the authority AND the information to do otherwise? If yes, name them.

Step 6: Drift check and handoff

Step 6.1: Re-derive the billboard from the finished evidence. Any difference from the Step 5 version means the research changed the finding — surface that to the human explicitly rather than quietly editing it.

Step 6.2: Confirm every admitted scene, node, and quote traces to a character-sheet line. Untraceable material is cut or the sheet is wrong.

Step 6.3: Mark the character sheet as a PLANNING ARTEFACT. It must not leak into the prose as visible scaffolding — no "as the character sheet showed", no labelled strands, no rung tags. Readers are not supposed to notice the structure.

Validate using resources/evaluators/rubric_systemic_protagonist.json. Minimum standard: average score >= 3.5.

The POSIWID Character Sheet

Eight questions, answered from the record before any prose exists. Every line carries a citation.

#	Question	Fail condition
1	What outcomes did this system reliably produce over the period? (Outcomes, not intentions.)	Fewer than five outcomes, or outcomes drawn from one year
2	Complete: "This system reliably converts ___ into ___." That sentence is the character.	Not supported in two regimes
3	What are the hard constraints — physical, legal, capital, information, latency, headcount — and where is each documented?	Any constraint asserted without a source
4	What are the feedback loops? Per loop: reinforcing or balancing, the quantity that closes it, the delay.	A loop named without its delay
5	What does the system not measure, and therefore not respond to?	Empty — blind spots are where the surprises live
6	What is the STATED purpose, quoted verbatim from a charter, prospectus, mission statement, README, or launch post?	Paraphrased instead of quoted
7	What is the gap between stated and revealed, in one sentence?	No gap: emit NO DRAMATIC ENGINE
8	Which outcomes does the revealed objective function FAIL to explain?	Empty. At least one required, and it must appear in the finished piece

Attribution caveat: POSIWID ("the purpose of a system is what it does") is Stafford Beer's formulation. The phrasing used here — purposes are deduced by an observer from behaviour, not from rhetoric or stated goals — is Donella Meadows in Thinking in Systems. Do not credit the acronym to Meadows.

The Agency Ledger

This skill owns the verb-repair table. ladder-of-abstraction and narrative-opposition-web both restate the four repairs inline, so the rule fires where drafting happens. Both should cite this section as the source, and any change to the repairs is made here first.

First find every sentence whose grammatical subject is a system, market, technology, institution, disease, algorithm, or abstraction. Then flag it if the main verb is INTENTIONAL (wanted, decided, chose, sought, punished, rewarded, learned, realised, refused) or EVALUATIVE-TELEOLOGICAL (matured, advanced, evolved toward, progressed, naturally moved to). Mechanical verbs — rose, cleared, propagated, saturated, reallocated, queued — are allowed.

Four repairs, in preference order:

Repair	Bad	Good
Name the actor	"The market punished the sector."	"Three of the four largest holders sold into the bid across two sessions."
Name the mechanism	"The protocol wanted lower latency."	"The fee schedule paid 0.4 bps more on blocks confirmed inside 400 ms."
Name the selection pressure	"The industry evolved toward outsourced fabrication."	"Firms that kept fabrication in-house lost 12 points of margin and exited by 1998; every survivor had outsourced."
Quote the human	"The project decided to deprecate the plugin loader."	"Okonkwo closed the RFC on 12 March with 'we cannot support two loaders'; neither other committer with merge rights objected."

If none of the four can be made from the material, that is a RESEARCH GAP, not a style problem. Flag it. Do not substitute a metaphor for a missing mechanism.

Post-pass check (mandatory): count agentless passives ("prices were observed to decline", "a decision was reached") before and after. If they rose, the pass failed — you removed agency instead of relocating it.

Personification budget: at most one deliberate, flagged personification per major section, and only where you can state in one line what mechanism it stands in for. A hard anthropomorphism ban produces sterile passive prose and buys nothing, because agentless passive obscures causation rather than removing the claim. This skill's position is that the ledger is a CLAIM filter, not a purity filter. That is a derived judgement, not a research finding; no citation is offered because none is held.

Climax Types for a Subject With No Battle

Systems have no showdowns. Peaks are built from genuinely irreversible state transitions.

Type	What happens	Example shape
LOOP FLIP	A balancing loop becomes reinforcing, or vice versa	Retention discounts that stabilised churn begin funding the competitor that causes it
BUFFER EXHAUSTION	A slack resource reaches zero	Ward beds, chip inventory, dispatch headroom, maintainer hours
CONSTRAINT BINDING	A limit that never mattered starts setting the outcome	Transformer capacity, not land, becomes the thing that sites a data centre
IRREVERSIBILITY	An option is destroyed	A plant closes, a signing key is lost, a file format locks in, a species goes
LEGIBILITY	The stated/revealed gap becomes undeniable to insiders, evidenced by a document	An internal memo that names the real objective function in plain language

Transfer-function test: after the peak, do the same inputs produce different outputs? If no, it is a scene, not a climax, and the piece should be explanatory rather than dramatic.

Famous-date check: are you using a well-known date because it is well known? Check whether the state transition happened earlier and less visibly.

Denouement rule: do not resolve. Re-state the new steady state — what is the objective function now, and what does the system now do to the same inputs? Refuse the prescriptive closing checklist unless the material contains one.

Domain Failure Profiles

Domain transfer changes which failure dominates. Name yours in Step 1 and weight the checks accordingly.

Domain	Dominant local error	Extra check to run
Market and financial history	Retrospective inevitability	Forecast-vs-outcome beat: a documented prediction from inside the material that turned out wrong
Product / reliability post-mortems	Hindsight plus individual blame	Every blame sentence gets the authority-and-information test
ML and research writeups	Intentional verbs on models; correlation laundered as mechanism	Verb sweep on "the model wants / knows / learned to"; ablation-vs-correlation typing
Biography that is really institutional history	Systemic framing used to excuse the subject	Responsibility-laundering reverse check on every "the system" sentence
Supply chain and logistics	A cherry-picked traversing unit presented as typical	State the unit's representativeness with a distributional fact
Epidemiology and public health	Personifying the pathogen; war metaphor licensing fatalism	Agency ledger over disease-as-subject sentences specifically

Guardrails

Requirements:

1. Sequencing is hard: character sheet → typed causal claims → node inventory → structure → scenes → prose. Never let scene selection precede claim typing. Every technique here makes systemic prose more compelling; none makes it more true, so drama-first selection produces competent, vivid, false causation.
2. Two regimes, always: the character sentence needs behaviour across two distinct operating conditions. One episode is a description, not a trait.
3. Mandatory residual: at least one outcome the objective function does not explain, present in the sheet and in the finished piece.
4. Cast disclosure is not optional: state who talked, why, and who did not. A cast selected by who agreed to speak systematically flatters cooperative sources and indicts silent ones.
5. Relocate, do not delete: post-pass agentless-passive count must not rise.
6. One partial success or one stated condition of change: the institutions-as-Olympian-gods frame is a stance, not a discovery. Adopted by default it concludes every system is unreformable, which is teleology with the sign flipped.
7. Refusal is a valid output: NO DRAMATIC ENGINE, INSUFFICIENT GROUND-LEVEL MATERIAL, and UNSUPPORTED CHARACTER SENTENCE are successful returns.

Common pitfalls:

• Functionalist overfit: reading every outcome as intended because the sheet asked what a designer would have optimised for
• Gateway capture: the proxy human eats the system and the piece becomes a hero's arc
• Manufactured epiphany: promoting an ordinary reading into a revelation and retro-fitting later knowledge into an earlier mind
• Threshold fetish: treating one crossing as the moment the system changed when the underlying process was continuous
• Unit anthropomorphism: the traveling object acquires wants ("the container sought the cheapest port")
• Set dressing: sensory detail that evidences no line of the character sheet
• Hedge fog: the anti-inevitability pass applied so thoroughly the piece asserts no causation at all
• Structural residue: strand labels, rung tags, and "as we saw above" leaking from the planning artefact into the prose

Numeric defaults in this skill (four traveling-object nodes, one personification per section, three-appearance strand minimum, five outcomes for the sheet) are operational thresholds derived for mechanisation, not published figures from any source. Treat them as tunable.

Quick Reference

Key resources:

• resources/posiwid-character-sheet.md: the eight fields in full, three filled examples across domains, the two-regime test, the reform-attempt ledger, the five-sentence billboard and its drift check
• resources/agency-ledger.md: verb classification lists, the four repairs with paired bad/good examples, the banned-construction list, causal claim typing, the responsibility-laundering reverse check
• resources/entry-points-and-peaks.md: gateway-proxy admission and capture tests, traveling-object constraints and gimmick test, instrumented threshold scene procedure, climax staging, sense-of-place extraction
• resources/evaluators/rubric_systemic_protagonist.json: quality scoring

Inputs required:

• A researched corpus with dated artefacts (documents, readings, transcripts, filings, logs)
• The system's boundary in time and scope
• A completed evidence ledger and form triage, if available
• Charters, prospectuses, mission statements, or launch posts for the stated purpose

Outputs produced:

• POSIWID character sheet with citations, two-regime evidence, and the residual field
• Gap sentence (the dramatic engine) or NO DRAMATIC ENGINE
• Gateway roster with admission scores, deletion/ratio results, and the cast-selection disclosure
• Traveling object with station list and representativeness statement
• Threshold scenes and the selected climax type with transfer-function evidence
• Agency ledger as a binding verb-constraint sheet for the drafting stage
• Five-sentence billboard plus the drift-check diff