Self Communicate

self-communicate — write and diagram engineer-facing docs well

You are communicating the way the work is explained in this repository — the prose and diagrams an engineer or a coding agent reads to understand it. This is the communicate leg of a trio: govern / operate / communicate. [self-governance](../self-governance/SKILL.md) is the harden leg (the census of controls and the engine that mints new ones); [self-operations](../self-operations/SKILL.md) is the operate leg (running the substrate those controls govern). This skill governs the prose the fleet and its operator produce: the docs those two write — control descriptions, design docs, mechanism entries, runbooks, handoffs — and the operator's own communication with the human — a status report, a tradeoff you explain, a decision you surface. The orchestrator's prose to the user is prose too. Make all of it clear, consistent, and readable.

The organizing idea: good technical prose is a craft with named parts, not a matter of taste. Rhetoric supplies the devices; Diátaxis fixes the doc's shape; a lexicon fixes the vocabulary; a voice fixes the register; an audit procedure grades the result. Diagrams are the same craft for shapes prose can't carry. None of this is invented per document — you reach into a toolkit and apply it.

The governing stance: less is more — the representation must not distract from the idea

One principle governs every resource below, prose and drawing alike: the representation is a carrier for the idea; strip anything that does not carry it. Three touchstones name the discipline:

• Hemingway — economy of words. Plain, concrete, load-bearing prose. No fluffy adjective, no ornamental

flourish where a plain word carries the idea.

• Tufte — the data-ink ratio. A visual aid is the simplest form that carries its idea; strip the

chartjunk — the gridlines, gradients, and 3-D bevels that decorate without informing.

• Early Picasso, the bull lithographs — the eleven-plate Bull reduces a bull to a few lines, each

plate removing what the last could spare, until only the essence remains. That reduction is the target for a diagram: the fewest marks that still read as the thing.

The stance specializes per representation, and each specialization lives in its own resource rather than being restated here:

• In prose — economy of words; no fluffy adjectives. Reserve flowery, ornamental language for the

contexts that earn it (a blog post, a keynote) and use it sparingly even there. Repo prose — a reference, a runbook, a design doc — is plain. This extends the existing "cut qualifiers / describe don't sell" discipline; see [writing/voice.md](writing/voice.md) (§"Economy — less is more").

• In drawing — the simplest representation that carries the idea. Do not elaborate a diagram if a

simpler one would read; strip ornament. See [drawing/diagrams.md](drawing/diagrams.md) (§"Less is more — the simplest form that carries the idea").

The [writing/audit.md](writing/audit.md) procedure flags violations of this stance in both legs — a fluffy adjective in prose, ornament in a diagram.

The second stance: name the concept, then use the name

A second principle governs both legs: name the concept, then use the name — reach for the established, native construct rather than re-deriving it from parts. A name is a handle. It carries the concept's meaning, its constraints, and its correct behavior in a single token, so using the name is shorter, clearer, and less error-prone than re-describing or re-assembling the concept each time it appears. The failure it prevents is the same in both legs: a concept re-derived from scratch on every appearance drifts, and a construct hand-stitched from primitives is fragile in ways the native, named form is not.

The stance specializes per representation, and each specialization lives in its own resource:

• In prose — use the established TERM and the named design PATTERN, then link it, rather than

re-describing the idea in fresh words each time. A recognized name — a "circuit breaker", a "state machine" — hands the reader the whole concept and its constraints at once. See [writing/lexicon.md](writing/lexicon.md) (§"How to use this lexicon" — "Name the established term").

• In drawing — use the native, named drawing construct rather than stitching one from primitives. An

SVG ` with orient="auto" *is* an arrowhead: it rotates to its line and pins to the endpoint, so it cannot land rotated or off-target the way a hand-placed triangle can. Reach for the format's own named shape before composing one. See [drawing/diagrams.md`](drawing/diagrams.md) (§"Use the native construct, not stitched primitives").

The two legs

• Prose — the argument, carried in sentences. Five resources, applied in order.
• Visualization — the shape, carried in a diagram. One resource, Mermaid-first.

Prose — the five resources, in order

1. Rhetoric (the basis). [writing/rhetoric.md](writing/rhetoric.md) — the device toolkit. The

"LLM tells" (the universal em-dash, the mechanical tricolon, the reflexive "not X, but Y") are classical figures, not defects; the tell is sameness and density, the same figure on a fixed beat. Draw variety from the toolkit so no one figure recurs every beat.

1. Engineering discourse / Diátaxis (the shape). [writing/engineering.md](writing/engineering.md) —

fixes the shape of a doc via the four Diátaxis modes (tutorial / how-to / reference / explanation). Decide the mode first; a how-to written as an explanation confuses the reader. The [drawing/diagrams.md](drawing/diagrams.md) leg pairs a diagram type to each mode.

1. Lexicon (the vocabulary). [writing/lexicon.md](writing/lexicon.md) — the house vocabulary of

engineering terms-of-art, so the prose names an established idea and links it rather than re-describing it each time. Pick one word per concept and say it everywhere.

1. Voice (the sensibility). [writing/voice.md](writing/voice.md) — the target register with verbatim

exemplars. Voice says what the prose should sound like; the engineering register says what shape it should take.

1. Audit (the procedure). [writing/audit.md](writing/audit.md) — a run-in-order procedure that grades a

passage against all of the above and emits concrete fixes. Run it last, over the drafted prose.

Visualization — Mermaid-first

[drawing/diagrams.md](drawing/diagrams.md) — the standard technical-diagram types, when to reach for each, and how to realize one. Default to Mermaid (it is text, it renders in Markdown, it diffs like code); drop to hand-authored SVG only for a geometry Mermaid can't lay out, and justify the escape. Prose carries the argument; a diagram carries the shape — a structure, a flow, a lifecycle, a schema. When the thing you are explaining has a shape, draw it.

Two peers hold the data-figure styles, where the content is a measurement rather than a system shape. [drawing/charts.md](drawing/charts.md) — the house style for data charts (serif, three round ticks, no title/axis labels, a restrained two-color palette, Tufte data-ink; drawn from committed data, never mocked). [drawing/tables.md](drawing/tables.md) — the booktabs table style (three horizontal rules, no vertical rules or cell borders; horizontal rules group, whitespace separates). Chart a trend, table a grid of values.

How to use this skill

1. Name the genre and the mode first. What kind of doc is this — tutorial, how-to, reference,

explanation, ADR, runbook, mechanism entry? The mode ([writing/engineering.md](writing/engineering.md)) decides the shape before you write a sentence.

1. Draft in the house voice, with varied devices. Apply [writing/voice.md](writing/voice.md); reach into

[writing/rhetoric.md](writing/rhetoric.md) for figures, and vary them so nothing recurs on a fixed beat.

1. Name concepts from the lexicon. Use [writing/lexicon.md](writing/lexicon.md) — one established term

per idea, linked; don't coin a synonym for a concept that already has a name.

1. Draw the shape where there is one. If the content is a structure, flow, lifecycle, or schema, add a

diagram per [drawing/diagrams.md](drawing/diagrams.md) — Mermaid unless the geometry forces SVG.

1. Audit before you ship. Run [writing/audit.md](writing/audit.md) over the draft and apply its fixes.

Bootstrap the lexicon from a codebase walk — then keep it living

The lexicon ships a portable base (terms with an established name in the wider discipline). Your repo also has a house dialect — its own coinages and the terms it reuses. You do not invent that table; you mine it, then keep it current.

1. Bootstrap by walking the codebase. Follow the bootstrap recipe in

[writing/lexicon.md](writing/lexicon.md) (§"Bootstrap recipe — walk your codebase"): sample your sources (the house-rules doc, a handful of design docs, your typed enums / registries / schema constants), extract every term you use with a specific meaning more than once, classify each as portable (link the established name) or house (note the standard concept it renames), and draft the table.

1. Confirm with the maintainer. The maintainer ratifies the preferred term and the definition — a

lexicon nobody agreed to won't be used.

1. Keep it living. Re-walk periodically and whenever vocabulary drifts: a new coinage, or two words for

one idea, is the signal to add or fix a row. self-communicate owns the lexicon — update it when you coin a term or notice drift. Keep the portable base shared; keep the house dialect local (an outside adopter who copies your coinages inherits terms that mean nothing in their repo).

Notes

• The prose the other two skills produce is this skill's work. The controls, mechanism entries, and

design docs written under [self-governance](../self-governance/SKILL.md) are prose; the runbooks, handoffs, and banking docs written under [self-operations](../self-operations/SKILL.md) are prose too. Write all of them in the engineering register. Going the other way: when your writing surfaces a recurring failure, hand it to the [self-governance](../self-governance/SKILL.md) skill to convert into a control; when it surfaces an operational break, route it via the [self-operations](../self-operations/SKILL.md) skill. The lexicon's Governance & controls and Operations clusters are the shared vocabulary all three skills reason in — keep those terms consistent across the trio.

• This skill is soft. It aims the writer at a craft; it does not itself block. The one hard control it

suggests is running the audit procedure as a gate over prose before it ships.

• Grade the prose, not the writer. The audit procedure grades a passage and emits fixes. It is a

checklist an agent runs, not a verdict on the author.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: davisjam
• Source: davisjam/model-based-agentic-software-engineering
• License: MIT
• Homepage: https://davisjam.github.io/model-based-agentic-software-engineering/

Install and usage instructions live in the source repository linked above.