Voice and language
Calm, specific, on the side of adoption.
We write for business leaders first, and for engineering and IT and Security on the same page.
Security is how an organization says yes to agents with confidence, so we never use it as a reason to slow down or stop.
Lead with the outcomeSay what gets safer first.
"Stop high-risk actions before they execute" comes before "a hook evaluates each tool call".
Why before howDetail is proof.
Name the control point, evidence, or workflow once the reader knows what it protects.
Concrete actionsEvery claim names something.
A user, a risk, an action, or an outcome. Verify, control, investigate, stop, approve, reconstruct.
Scope and limitsSay where it stops.
State what a check covers and what evidence it rests on. "Browser results are an advisory preview" is on-brand.
Short and directOne idea per sentence.
Direct verbs, plain words. Cut anything that does not change the meaning.
Calm confidenceNo shouting, no scaring.
Specific and practical. We never scold the reader or sell with fear.
The headline pattern
Two short sentences. A lead that states the benefit, then an accent in a contrasting color that adds timing or proof.
Let every team use AI. Keep agent risk under control.
Find every AI agent on every device. Check every skill it loads.
Find the signal in your AI logs. Follow every action.
We say, we avoid
Reduce agent risk without slowing down the teams using AI.Adoption and control in one line.
Lock down rogue AI before it's too late.Fear framing; treats AI as the enemy.
Stop high-risk actions before they execute.Concrete action and timing.
Seamless, next-generation agent protection.Filler that names no risk or outcome.
Find risky skills, MCP servers, plugins, and packages before teams trust them.Scoped to named components.
Complete security for every AI threat.Unscoped absolute with no evidence.
Investigate incidents with the evidence already attached.Shows the benefit to the analyst.
Security, not friction.Rhetorical contrast formula.
Supply Chain, Runtime, and Hunt.Exact names, lifecycle order.
Our arsenal of AI defense weapons.War metaphor. Our name refers to signals.
Naming
| Use | Avoid |
| Signals Corps | Signals, Signal Corps, SignalsCorps, SC |
| Supply Chain, always | Supply Chain Security, SCS, supply-chain |
| Runtime | Runtime Protection, RTP, runtime guard |
| Hunt | Signals Hunt, the Hunt product |
| Module: the Supply Chain module, Control Plane modules | product, app, tool, feature, solution, add-on |
| Signals Corps Control Plane, then Control Plane | the Signals Corps app, the app, the dashboard, the console, the suite, the solution |
| AI agents, then agents | bots |
Words we use
- Visibility, governance, and control
- What we give IT and Security.
- Agent risk
- What we reduce. Avoid "eliminate".
- Control point
- Where a decision happens: hook, gateway, proxy, extension, endpoint agent, or API.
- Allow, warn, request approval, block
- The four runtime outcomes, in this order.
- Evidence already attached
- Rule, content, decision, and outcome kept together.
Platform and module names
Signals Corps Control Plane is the platform. It hosts the modules: Supply Chain, Runtime, and Hunt. It lives at controlplane.signalscorps.com and opens to early sign ups in the next few weeks. Until then, link calls to action to the early access waitlist. Write "Signals Corps Control Plane" on first mention and "Control Plane" after that. Never call it "the Signals Corps app", "the app", "the dashboard", or "the console". Supply Chain, Runtime, and Hunt are always called modules ("the Supply Chain module", "Control Plane modules"), never products, apps, tools, features, solutions, or add-ons. When the relationship matters, describe a module as part of Control Plane: "Hunt, in Signals Corps Control Plane, searches agent sessions and tool calls."
Words we leave out
seamlessnext-generationcutting-edgeholisticrobustunlockempowersuperchargegame-changingbulletproof100% securerogue AIarsenalwar room
Mechanics
How a Signals Corps sentence is built, for everything we write: website pages, docs, blog posts, emails, decks, social posts, and support replies. Write this way from the first draft.
Explain clearly
Start with what the reader gets
Say what changes for them, then how it works.
Runtime stops a risky command before it runs. Then: a hook sends the command to the policy engine.
Give an example for every claim
Add the detail that makes it ours: the file, command, rule, or event.
Runtime stopped curl -T .env to an unapproved domain and showed the developer the rule that matched.
Show the real thing
Commands, paths, config keys, rule IDs, and log lines go in code formatting so readers can copy them.
Define terms once
The first time a page says "MCP server" or "tool poisoning", explain it in plain words. Then use the same term every time.
Show your working
On every page and in every format: name what was tested and the version, whether a claim comes from observed behavior, synthetic examples, or documentation, and where the check stops.
No binary or version inspected.
Match instructions to the risk
Name the risk and its consequence, then what to do. Keep "must" and "never" for real requirements, with the reason. No scolding.
Write like a person
Looks generated
Write instead
It's not a scanner, it's a trust layer.Not X but Y
Supply Chain finds the AI agents on each device and checks the skills they load.State the point directly.
No guesswork. No blind spots. Just control.One-line closers and fragments
End on the last concrete fact.Cut lines that only add emphasis.
Trust is the currency of agentic AI.A saying that sounds deep
Teams adopt agent components faster when they can see what each one is allowed to do.Replace the saying with the claim.
Here's the thing: agents need guardrails. Let's dive in.Staged run-up
Agents can run shell commands with your credentials.Start with the point.
Faster, safer, smarter agent adoption.Forced three
Teams adopt new agent tools without waiting for a manual security review.List three only when there are three.
Risky actions are blocked automatically.Hidden actor
Runtime blocks the action and shows the developer why.Say who does what.
Hunt serves as an investigation surface that boasts timelines.Avoiding "is" and "has"
Hunt is where analysts investigate. It has session timelines.Use the simple verb.
Experts agree agent security is critical.Borrowed authority
Name the source and what it said, or cut the sentence.Never invent a source.
Format
American English
Organization, behavior, analyze, color, center, catalog, program.
Sentence case
Headings, labels, and buttons. Buttons start with a verb: "Get early access".
Punctuation
No em or en dashes. Use a comma, colon, parentheses, or a full stop. Straight quotes and the Oxford comma.
Plain formatting
Bold only what readers must not miss. No bold label on every list item, and no emojis or arrows in headings.
Words that usually mean generated text: delve, crucial, pivotal, robust, landscape, showcase, underscore, foster, testament, vibrant, seamless, next-generation, game-changing, unlock, empower. Use the plain word. Mockups and dashboards with invented numbers carry the label "Illustrative data".