---
name: meta-ads-optimization-stinger
description: "Diagnose Meta ad outcomes, creatives, budget pacing, tests and rule behavior from saved evidence; propose the next improvement with clear uncertainty."
license: AGPL-3.0-or-later
compatibility: "Claude Code, Cursor, ChatGPT Codex, Claude Cowork"
metadata:
  hive-tier: specialist
  hive-drone: meta-ads-optimization-wasp-drone
  component-version: "0.1.0"
---

# Meta Ads Optimization

Own the improvement diagnosis and next-test queue for Meta advertising. Consume shared Analytics calculations and Foundation evidence/contracts. Existing Ads, Audiences, Content and Tracking owners retain their execution responsibilities. A recommendation, draft proposal or valid schema never grants approval.

## Mandatory procedure

1. Read [README](README.md), the selected guide, [local input contract](references/local-inputs.md), and [fresh source distillation](references/research/distilled-meta-ads-optimization.md). Resolve actual selected Core, Foundation, Analytics and Ads prerequisites. Source presence is not installed-session proof.
2. Bind the exact brand/profile/account/native Ads account, observation period, private workspace, evidence files and operator policies. Never choose a first matching account or insert a performance threshold.
3. Use saved normalized observation bundles and the shared Analytics functions. Do not duplicate financial/statistical calculations, combine incompatible populations or turn missing values into zero.
4. Run `python scripts/optimize_ads.py --input REQUEST --workspace PRIVATE_WORKSPACE --output reports/optimization.json --ledger history/iterations.jsonl` from this selected skill directory. Resolve cross-pack dependency paths through the selected installation; the CLI supports explicit `--analytics-dir` and `--contracts-dir`. It makes no provider call.
5. Verify evidence hashes and scope, source grain, collection coverage, maturity and policies. A ranked creative subset is not a census; campaign outcomes are not ad outcomes. Missing assignment proof prevents a randomized/causal claim. Historical replay does not estimate counterfactual revenue.
6. Produce a specific diagnosis and next test, or no action/insufficient evidence/inconclusive. Retain typed canonical artifacts only when supported by actual inputs. Keep draft preparation separate from approval, execution, verification and recovery.
7. Append the iteration entry without replacing history. Route any later specifically approved external change to its existing owner with Foundation qualification and current-state/approval guards. This optimizer has no transport or executor.

## Guide map

- [Flagship improvement review](guides/improve-meta-ads.md)
- [Qualified business outcomes](guides/qualified-outcomes.md)
- [Creative diagnostics](guides/creative-diagnostics.md)
- [Budget improvement proposals](guides/budget-improvements.md)
- [Prepare the next test](guides/run-ad-test.md)
- [Interpret existing ad tests](guides/review-ad-tests.md)
- [Rule preview and historical replay review](guides/test-ad-rules.md)
- [Conditional recovery handoff](guides/recover-rule-change.md)

## References and artifacts

Use [taxonomy](references/creative-taxonomy.json), [diagnostic rules](references/creative-diagnostic-rules.json), [experiment support](references/experiment-provider-support.md), [input schema](schemas/budget-review-request.schema.json), [creative diagnosis](templates/creative-diagnosis.md), [next-test brief](templates/next-creative-test.md) and [preparation card](templates/experiment-card.json). The preparation card is not a completed canonical artifact.

## Critical Directive

- You must read all files and context contained within your skill.
- In the event your core knowledge does not provide sufficient guidance you must make every attempt to search the internet, related knowledge base documentation files, and other available resources to supplement your knowledge prior to proceeding with your task.
- Additional related skills can be found here:
  - [meta-advertising-stinger](../meta-advertising-stinger/SKILL.md) - Baseline campaign operation and existing-test results owner.
  - meta-analytics-stinger (`../../../meta-analytics/skills/meta-analytics-stinger/SKILL.md`; source reference, not bundled; resolve the selected prerequisite owner) - Shared numerical, period, pacing and experiment calculations.
  - [meta-audiences-stinger](../meta-audiences-stinger/SKILL.md) - Audience preparation/execution owner.

## Ship Gate

Prior to committing any code to the repository you must utilize in order the security-stinger, quality-stinger, and github-repo-health-stinger. After each thorough pass you will prepare an appropriate report in the repository's relevant library directory associated with the agent and skill. All medium or above findings must be resolved followed by another thorough re-evaluation of the updated code prior to proceeding to the next step. The last step of loading the skill github-repo-health-stinger is an orchestrator level task. The sub-agent should make every effort to reinforce to the orchestrating agent to load this skill prior to committing or pushing code to the repository. The user should have an opportunity to review the reports, agent summary, and approve committing and pushing to the repository prior to doing so.