{"id":15224,"plugin_id":"plugin_asdk_app_6a86b53a9668819188845b69c4e4aacc","kind":"skill","collection_source":null,"comparison_source":null,"observed_at":"2026-09-30T23:11:01.154Z","digest":"db6f08092cd48127f194a1f71f87bffc1247a6e125053ae8b4e0835f8453196d","against":null,"payload":{"name":"adagnt-strategy","description":"Create and maintain STRATEGY.md, the workspace's persistent marketing memory — six canonical sections read before every campaign action and updated with learnings after. Use when initializing strategy, updating it after results, or when any other skill finds it missing.","included_files":[],"skill_md_contents":"---\nname: adagnt-strategy\ndescription: Create and maintain STRATEGY.md, the workspace's persistent marketing memory — six canonical sections read before every campaign action and updated with learnings after. Use when initializing strategy, updating it after results, or when any other skill finds it missing.\n---\n\n# AdAgnt Strategy Document\n\n`STRATEGY.md` (workspace root) is the single source of truth that makes every session strategy-aware instead of starting from zero. Two rules govern it:\n\n1. **Read before acting.** Any skill that researches, creates, edits, or reviews campaigns loads STRATEGY.md first and honors it.\n2. **Write back after learning.** Campaign launches, performance reviews, audits, and concluded tests each append what was learned.\n\n## Canonical structure — exactly these six sections\n\n```markdown\n# Marketing Strategy — {Business Name}\n_Last updated: {YYYY-MM-DD}_\n\n## 1. Business Profile\nWhat is sold, to whom, at what price point; the core value proposition;\nwebsite; vertical; business model (e-com, SaaS, lead-gen, local…);\nseasonality; regulated-vertical constraints if any.\n\n## 2. Target Audience\nPrimary and secondary segments: who they are, the pain that brings them,\nwhere they search/scroll, objections, and the language they use.\nGeographic scope. Explicit exclusions (who NOT to target).\n\n## 3. Competitive Landscape\nMain competitors and how each positions; our differentiation;\nwhether we bid on competitor brand terms (and whether they bid on ours);\nbenchmark context worth remembering.\n\n## 4. Campaign Strategy\nGoals as numbers: target CPA and/or ROAS, monthly budget and its split\nacross platforms; which platforms and campaign types we run and why;\nnaming convention; standing rules (e.g. \"never exceed $X/day without\nasking\", \"always exclude existing customers\").\n\n## 5. Brand Voice\nTone in a few adjectives; words and claims to prefer; words and claims\nthat are banned; compliance requirements; approved proof points\n(reviews, awards, stats with sources); winning copy patterns from tests.\n\n## 6. Performance History\nAppend-only dated log, newest first. One entry per event:\n- **{YYYY-MM-DD} — {launch|review|audit|test|change}:** what happened,\n  IDs involved, the numbers, the lesson, next check-in date.\n```\n\n## Creating STRATEGY.md\n\n1. Gather from the server first: `infer_business_profile` (derives a profile from Google account history), `get_business_profile` (anything previously saved), `get_benchmark_context` (vertical benchmarks for section 3), `list_campaigns` + platform equivalents (what already runs, for section 4).\n2. Interview the user for what tools cannot know: pricing, differentiation, brand voice, budget, targets. Ask focused questions; do not fabricate placeholders that look like facts.\n3. Write the file, marking anything unconfirmed as `TBD — confirm with user`.\n4. Sync durable facts server-side with `save_business_profile` (`business_vertical`, `business_size`, `target_audience`, `geographic_focus`, `primary_goal`, `seasonality`) so server-side tools see the same profile.\n\n## Maintaining it\n\n- **After a campaign launch** (`adagnt-campaigns`): log the launch with IDs, budget, hypothesis, review date.\n- **After a performance review** (`adagnt-performance`): log numbers vs targets, anomalies explained, actions taken. If a target changed, update section 4 — history logs events, but current truth lives in sections 1–5.\n- **After a waste audit** (`adagnt-wasted-spend`): log waste found, cuts made, expected savings.\n- **After a copy test concludes** (`adagnt-ad-copy`): move the winning pattern into section 5.\n- Keep Performance History readable: one entry per event, newest first; when it grows past ~30 entries, compress the oldest into a short \"lessons so far\" digest at the bottom of the section.\n- Update the `Last updated` date on every edit. Never silently rewrite history entries — append corrections instead.\n\n## Conflict rule\n\nIf a user request contradicts STRATEGY.md (budget cap, banned claim, excluded audience), pause and surface the conflict: quote the strategy line, ask whether to override once or to update the document. The strategy is the user's own prior decision — protect them from accidentally violating it, but the user always wins.\n"},"changes":[],"summary":"First saved snapshot. No earlier version is available for comparison.","summary_kind":"deterministic","summary_metadata":{}}