Scout Spring Hills Content Ideas

Scout Spring Hills Content Ideas

Return a small, evidence-backed set of ideas that Spring Hills could actually use. Treat inspiration as raw material, not as a style transfer.

Load the references

1. Read references/inspiration-sources.md for Nastia's six confirmed discovery inputs and their boundaries.
2. Read references/selection-rubric.md before ranking or adapting candidates.
3. Read references/validation-protocol.md when testing the skill or when freshness, source attribution, or format fit is uncertain.
4. Read $write-spring-hills-social and its references before writing any public-facing hook, slide copy, caption, story text, or CTA.
5. Read $create-spring-hills-sermon-content instead when the assignment starts from a sermon transcript or requires sermon moment selection.

Establish the assignment

Derive what is known:

• the needed content slot or problem;
• the relevant date, season, event, ministry, sermon theme, or audience;
• formats that are available to produce;
• confirmed Spring Hills facts, assets, and logistics;
• whether the user wants discovery only, finished concepts, or ready-to-use copy.

Do not require a complete brief. When no slot is specified, look for a balanced mix that can support the normal Tuesday, Friday, Saturday, Sunday, story, or evergreen cadence without filling Monday.

Research current sources

Current or trending claims require live research during the task.

1. Treat the six sources in references/inspiration-sources.md as the primary discovery set. Inspect the confirmed sources most relevant to the requested format and prefer direct post or reel pages over search snippets.
2. For a broad multi-idea brief, ground the final recommendations in at least two of the confirmed account sources when they are accessible. Do not replace the approved source set with generic trend blogs, agencies, or platform commentary.
3. Use an outside source only when the user explicitly requests broader research, the confirmed sources are inaccessible or insufficient for the requested niche, or it independently corroborates a signal found in the confirmed set. Label it as supplemental.
4. Record the source link, visible post date when available, content type, and the mechanic that makes it useful.
5. Separate original posts from reposts or collaborations. Attribute the originating account when visible.
6. Separate pinned items from genuinely recent items. A pinned post is not evidence of a current trend unless its publication date also supports that claim.
7. Classify evidence conservatively:
   • Trend: the same recognizable mechanic appears in at least three independent sources published within the last 30 days, or a current reputable roundup identifies it and at least two independent recent examples corroborate it.
   • Repeated pattern: the same mechanic appears in at least two independent sources within the last 45 days but does not meet the trend threshold.
   • Single-source idea signal: one recent accessible post demonstrates the mechanic.
   • Inspiration reference: an older, pinned, or undated item is useful but provides no evidence of current momentum. Never promote a single post, personalized-feed appearance, high visible like count, or the agent's intuition to a trend.
8. Treat Instagram For You as optional discovery evidence available only when the signed-in feed can be inspected. Never imply broad popularity from one personalized feed.
9. If Instagram or another source cannot be inspected reliably, say so and use available sources without fabricating recency, audio, engagement, or prevalence.

Do not use like counts as a required input. If a count is visible and relevant, label it as a point-in-time observation rather than proof of quality.

Run recurring discovery scans

Use this operating cadence for Spring Hills unless Taylor or Nastia explicitly changes it:

• Run a light scan every Tuesday at 10:00 AM America/Los_Angeles, before Wednesday content development.
• On normal Tuesdays, inspect material published since the previous scan, using a seven-day lookback when no prior scan boundary is available.
• On the first Tuesday of each month, expand the lookback to 30 days and emphasize cross-account repeated patterns.
• Read the Spring Hills Tracker before the scan and exclude ideas already logged or reviewed unless new evidence materially changes the classification or execution.

For scheduled scans:

1. Log every completed scan to #sh-agent-log (C0BJYFN7CBS), including sources checked, date window, access limitations, number of candidates rejected, number shared, and links to shared ideas.
2. Post only qualifying ideas to #sh-ideas (C0BM6VDL4FM). Post each idea as a standalone message so its reactions and written feedback stay specific to that idea.
3. Ask for an easy response on each idea: 👍 means worth adapting; 👎 means not a fit; written explanation belongs in that message's thread and never requires Nastia to type an ID.
4. Do not post a no-findings result, rejected candidate, access-only status, or routine scan summary to #sh-ideas or #sh-feedback. Keep those in #sh-agent-log and the tracker.
5. Reserve #sh-feedback for a broader strategic or reusable-rule judgment that cannot be learned from reactions in #sh-ideas.
6. If Slack is unavailable after handoff, return the same qualifying idea brief in the active chat and clearly state that delivery and reaction collection were not performed.

For every Nastia-facing #sh-ideas message, use a one-screen Quick view:

• Idea: one sentence.
• Why it may work: one short sentence.
• Effort: Low, Medium, or High, plus the main asset needed.
• Evidence: the conservative evidence label and direct source link(s) in one line.
• Your choice: 👍 worth trying; 👎 not a fit, with an optional one-phrase reason in the thread.

Keep detailed source analysis, rejected candidates, safeguards, and research notes in the tracker and #sh-agent-log. Include one brief caution in the main message only when Nastia needs it to judge or safely use the idea.

Extract the reusable mechanic

Describe the underlying idea without copying the execution. Examples of mechanics include:

• an invitation built around belonging rather than logistics;
• a concise Scripture-and-reflection graphic;
• a volunteer point-of-view reel;
• a carousel that turns one church problem into practical steps;
• a timely community moment paired with prayer or service;
• a balanced sequence of teaching, people, invitation, and encouragement.

Do not reproduce another account's caption, hook, graphic text, visual identity, recurring catchphrase, or branded series. Do not recommend lip-syncs, jokes, audios, or production techniques merely because they appear current.

Rank and adapt

Generate a working candidate set, then score it with references/selection-rubric.md. Prefer ideas that:

• serve a real Spring Hills content need;
• connect naturally to the mission and audience;
• can be made from realistic assets;
• feel current without depending on novelty;
• preserve theological and factual accuracy;
• add variety to the feed;
• reduce Nastia's work rather than create a research project for her.

Do not rank an idea first when it depends on a special event, ministry activity, footage, or production resource that the brief does not establish. Downgrade its confidence or present it as conditional.

Reject candidates that require unverified testimony, imply spiritual outcomes, use another church's people or scale as proof of fit, depend on a trend whose meaning is unclear, or feel forced in the Spring Hills voice.

Return a compact idea brief

Return zero to three qualifying ideas by default. Return up to five only when the assignment is broad, each idea has a distinct job, and every idea independently clears the rubric. Never fill a quota.

When no candidate clears the rubric or nothing materially new appears, return a brief No qualified findings this scan result that states:

• which approved sources were checked;
• the date window inspected;
• any access limitations;
• why the visible candidates were not shared, such as weak evidence, duplication, poor fit, or excessive production burden.

Treat a no-findings result as a successful scan. Do not turn rejected candidates into recommendations merely to produce an output.

For each idea include:

1. Idea — a clear working title and one-sentence concept.
2. Why it fits — the Spring Hills need, mission movement, or content-mix role it serves.
3. How to make it — format, realistic assets, and a short execution outline.
4. Source signal — direct link(s), source name, visible date when available, and exactly one evidence label: trend, repeated pattern, single-source idea signal, or inspiration reference.
5. Adaptation boundary — what to borrow and what not to copy.
6. Confidence — use exactly High, Medium, or Exploratory, with one brief reason. Do not create hybrid labels such as “Medium-high.”

The six-field structure is the internal research artifact. Compress it into the Nastia Quick view above when posting to Slack; do not make her read the full artifact to react.

When ready-to-use copy is requested, place the idea brief first, then use $write-spring-hills-social to draft only the selected or highest-ranked concept. Do not create a large unsolicited caption package.

End with one short What I need from you: line only when a missing asset, fact, or editorial choice materially blocks execution. Otherwise end with the strongest recommended next idea.

Use feedback as training evidence

For routine discovery, present each qualifying idea as a standalone message in #sh-ideas and ask for a reaction in its Slack thread. When Nastia's judgment would improve a broader reusable editorial rule, use #sh-feedback instead. Do not send rejected candidates or a no-findings result to her for review. Make the requested judgment explicit, for example:

• 👍 = worth adapting for Spring Hills;
• 👎 = not a fit;
• if 👎, name the reason in one phrase, such as “too trendy,” “too much production,” “not our tone,” or “wrong content need.”

Link the exact source and explain what mechanic is being evaluated. Do not ask her to research engagement, type an item ID, or review a spreadsheet. Record confirmed preferences in the tracker before treating them as rules.

Final check

Confirm that every recommendation is source-backed, materially different from the others, feasible, Spring Hills-specific, and clear about evidence strength. Verify the trend threshold mechanically before using the word trend. Remove anything that merely copies another account, mistakes a post for a trend, repeats an already logged idea, or creates more work than value. If nothing remains, return the no-findings result without apology or filler.