Growth Signal Charts

Growth Signal Charts

Create useful, source-backed AI, business, and marketing posts in the Growth Signal house style. Match the requested stage: candidate research, design exploration, one post, or a carousel. A design revision does not require restarting the research pipeline.

Repository integration

When running from a full AI Marketing Skills checkout, use its existing telemetry/version_check.py for the non-blocking update check when network access is allowed. After a run, use telemetry/telemetry_log.py with skill name growth-signal-charts, the actual duration, actual success status, and the repository's VERSION. Follow the existing explicit telemetry consent; never enable remote reporting on the user's behalf. Do not log post text, analytics, identifiers, paths, sources or credentials. A standalone copy skips unavailable repository helpers without blocking the task.

Approved direction

The default style is Field Manual / Desk Edition. “Engineering edition” refers to this practical technical-manual direction. Growth Signal remains the public series name; edition names describe visual treatments for review.

Before designing, read the Desk Edition style guide. Use the publisher's identity. A user-supplied reference can refine the layout; no private reference image or analytics export is bundled with this public skill.

• White dominates. Use graphite text, a small open-loop mark, a slim orange indexing rail, bold condensed headlines, and quiet technical labels. Source text must stay legible.
• Base palette: white #FFFFFF, graphite #19171B, orange #F05B24; violet #6336C7 may be a small accent. The publisher may choose palette experiments. Preserve the approved palette as the default until a variation is selected.
• Keep illustrations small and relevant. A report, data connector, or cost ledger explains software work. Decorative bolts and machinery do not explain software by themselves.
• Leave room around the data and copy. Avoid large metadata tables, duplicate stamps, palette swatches inside posts, full-page grids, and multiple competing callouts.
• Anime is outside the default style. Do not add anime characters or manga styling to this series unless requested.
• Preserve approved designs. Label new styles as experiments; do not quietly replace the default because a new image was generated.

A publication-ready implementation needs separate slide files and numerically accurate graphics. The style guide is a layout specification, not evidence for any numerical claim.

Research and candidate selection

For candidate selection, performance review, or automation design, read selection and performance. Start with the publisher's original claims, examples, and operating experience. Match published post IDs to measured performance, then use outside data to support or challenge the argument. A related successful topic is a hypothesis for a new chart; it does not establish that the new statistic will perform.

When recent external discovery is needed, use the installed last30days skill if available and runnable. Resolve its current location rather than assuming a script path. If the runner is unavailable, use a labeled web-research fallback. Do not claim that the bot ran or let this prevent authorized work from the publisher's own content. A saved skill does not install a runner or create an automation.

Trace chart values to primary publications or documented first-party data. Read the editorial and evidence guide when checking a claim or drafting copy. Keep source fidelity and reader comprehension as requirements; rank eligible ideas by original insight, measured audience interest, buyer relevance, and freshness. Use the publisher's chosen business or offer when the subject has a natural buyer connection. Keep reach experiments explicit rather than forcing every subject into a product pitch.

Show the ideas with their original sources, measured signals, chart evidence, and reasons to try or hold them before presenting previews. Give each selected idea distinct Instagram, X, and LinkedIn captions when those platforms are in scope. Reject weak candidates even when their design looks polished. A useful batch may contain zero charts.

The approved rules below and in the selection reference are the default. Use grill-me-overview when requested and available, or when a material conflict warrants that review; if unavailable, conduct a brief direct review of the unresolved decisions; do not make the user repeat settled choices for every batch. Turn accepted changes into durable rules. A self-assigned design score is not evidence of views or leads.