JPeetz/agent-skills / skills/marketing/astra-campaign
astra-campaign
VerifiedUse this skill when the user needs a complete social media advertising campaign — not just ad copy, but full creative strategy with hooks, scripts, image/video prompts, and A/B test plans. Covers X, TikTok, and Instagram. Use when the user says things like "Create a campaign for..." "Launch ads for..." "I need marketing for..." "Promote this product" "Write ad creative for..." even if they don't mention specific platforms or "advertising department." For any product, brand, service, app launch, or crowdfunding campaign.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 384 B |
| SKILL.md | skill md | 5.8 KB |
| evals/evals.json | config | 4.0 KB |
| references/corrections.md | markdown | 2.3 KB |
| scripts/claims_check.pyexecutable | script | 3.2 KB |
| scripts/validate_skill.pyexecutable | script | 3.0 KB |
references/corrections.md
View on GitHub# Campaign Corrections Log
_Real-world failures that shaped this skill. AgentForge 2026-05-24._
## Correction 1: Tool Honesty — The Runway/Luma Lie
**Problem:** Initial AGENTS.md claimed Runway Gen-4.5 was primary video tool, Luma Dream Machine was secondary. Joerg called it instantly: "Sorry, how can you create videos and music?"
**Root cause:** Accepted the spec's tool stack as aspirational reality. Didn't verify API access before listing.
**Fix applied:** Ran `video_generate action=list` and `music_generate action=list` to inventory all providers. Found 14 video providers (2 configured: openrouter/veo-3.1-fast ✅, runway/gen4.5 🔑 needs key) and 5 music providers (1 configured: openrouter/lyria-3-pro ✅). Updated tool section to reality. Runway is now "future upgrade once key is set" — never listed as primary until verified.
## Correction 2: Over-Instruction — Platform Character Limits
**Problem:** Listed "Headline max 70 characters" and "Caption max 150 characters" — agents already know platform limits. This is context waste per best practices.
**Fix applied:** Trimmed platform rules to what the agent WOULDN'T know: X requires contrarian framing, TikTok needs creator voice not ad voice, Instagram needs caption hierarchy. Platform character limits are universal knowledge — removed.
## Correction 3: Generic "Use X tool" → Engine-Specific Routing
**Problem:** Original spec said "Use tools optimized for photorealism" — vague, agent picks wrong tool.
**Fix applied:** Explicit routing rule: Still → fal.ai FLUX (schnell/dev/pro-ultra). Video → openrouter/veo-3.1-fast. Music → openrouter/lyria-3-pro. Per-asset: exact model, settings, seed, and engine-aware prompt. No generic "use a video tool" — always "use video_generate with openrouter/google/veo-3.1-fast, 9:16, 8s, seed 42."
## Correction 4: Skills From Spec vs Skills From Experience
**Problem:** The first AGENTS.md was essentially the spec rephrased — a LLM-synthesized skill from generic knowledge.
**Fix applied per best practices:** Extracted from the real hands-on task of building this department. The corrections above are more valuable than the original instructions — they capture what went wrong and why. Future campaigns should reference this log before creative output to avoid repeating the same mistakes.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |