oegeyilmaz9/seo-aeo-geo-ultimate / skills/seo-action-plan
seo-action-plan
VerifiedUse when turning validated SEO, AEO, GEO, technical, or measurement findings into an approval-ready, evidence-linked action plan without claiming guaranteed ranking, citation, traffic, or conversion outcomes.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 7.9 KB |
| agents/openai.yaml | config | 405 B |
| references/action-plan-protocol.md | markdown | 5.1 KB |
| references/contracts/action-plan.schema.json | config | 10.2 KB |
| references/contracts/contracts-lock.json | config | 2.5 KB |
| references/contracts/evidence-record.schema.json | config | 1.7 KB |
| references/contracts/optimization-brief.schema.json | config | 6.3 KB |
| references/contracts/research-pack.schema.json | config | 5.2 KB |
| references/contracts/seo-findings.schema.json | config | 5.2 KB |
| references/markdown-delivery-contract.md | markdown | 4.3 KB |
references/action-plan-protocol.md
View on GitHub# Action-plan protocol
## Intended use
`action-plan.json` is an approval and handoff artifact. It converts a validated AEO/GEO optimization brief or conventional `seo-findings.json` bundle into discrete actions without changing the source site. It is deliberately not a backlog export, a universal SEO score, or proof that a platform will retrieve, cite, rank, or recommend a page.
The artifact is valid only with immutable upstream input bundles under `inputs/`. An AEO/GEO input bundle contains the original `optimization-brief.json`, `research-pack.json`, captures, metadata, and raw evidence required by its producer’s validator. A conventional SEO input bundle contains the original `seo-findings.json` and every `raw_evidence_ref` named by it.
## Bundle shape
```text
action-plan-bundle/
action-plan.json
inputs/
aeo-homepage/
optimization-brief.json
research-pack.json
raw/...
technical-homepage/
seo-findings.json
raw/...
```
`input_briefs[].bundle_ref`, `input_findings[].bundle_ref`, and their `artifact_ref` values are relative to the action-plan bundle. They cannot escape it or traverse a symbolic link/reparse point. The validator hashes each artifact, resolves its input bundle, and reruns its declared upstream validator. A plan using SEO Findings `1.1.0`, Optimization Brief `1.1.0`, or expanded action/verification types must use Action Plan `1.2.0`; legacy Findings `1.0.0` may use Action Plan `1.1.0`, and a legacy Optimization Brief `1.0.0`-only plan remains supported.
## Required action logic
Every action must carry all of the following.
| Field | Requirement |
| --- | --- |
| `finding_refs` | At least one finding from one input brief or SEO finding set. |
| `evidence_refs` | At least one resolving evidence record; it must overlap a linked finding’s evidence. |
| `owner` | A real responsible team/role and a distinct approval role. The planner cannot own implementation. |
| `execution_outline` | Concrete, bounded steps. State what will not be changed where ambiguity matters. |
| `acceptance_criteria` | Reviewable conditions, not uplift or citation promises. |
| `verification` | A method, source-specific metric type, success condition, and guardrail. |
| `rollback` | A required/reversible path for potentially harmful change. |
| `claim_boundary` | A scope note plus `no_guarantee: true`. |
Use `confidence: high` only when all linked finding and evidence classifications are `confirmed`. `medium` is allowed only with `confirmed` or `vendor-recommended` support. `low` is limited to an explicitly monitored experiment and stays pending approval. No action can be based on a speculative finding or evidence item.
## Priority and risk
`now` means an approval-ready, evidence-backed action proposed with an owner, acceptance criteria, and rollback; its formal approval status is still `pending`. `next` is similarly prepared but sequenced behind `now`. `later` is valid but should name the dependency or missing decision. `do-not-do` belongs in `declined_actions`, not `actions`.
Risk is operational risk, not a prediction of search impact. Action Plan `1.2.0` requires an explicit `risk_flags` array; declare every applicable `policy`, `crawler-policy`, `indexing`, `legal`, `privacy`, `accessibility`, `experiment-wide`, or `irreversible` flag. Any declared flag requires `risk: high`, an explicit rollback method, and an appropriate approval role. A linked policy, accessibility, or documented crawler-control finding also requires its matching flag and remains high risk even when the proposed implementation is described as technical or content work. The validator can enforce declared and linked risk, but the reviewer remains responsible for rejecting an omitted semantic risk.
## Measurement rules
Choose the metric that actually observes the proposed outcome:
- `recrawl` for rendered/discovery verification;
- `manual-review` for factual, editorial, or accessibility acceptance;
- `search-console` for Google Search reporting;
- `analytics` for referral or conversion events;
- `visibility-run` for a frozen prompt/corpus observation run;
- `seo-performance-run` for a hash-pinned conventional search comparison;
- `log-review` for crawler requests only;
- `merchant-feed`, `business-profile`, `video-indexing`, `task-flow`, or `site-graph` for their matching specialist evidence.
Keep these measures separate. For example, a bot request can support `log-review`, but it cannot establish a citation or referral result. A visibility run is observational and needs comparable repeated samples before describing drift.
## Declining a request
Add a `declined_actions` entry when a requested change would make an unsupported platform claim, rely on stale/absent evidence, create policy/legal/accessibility risk, or confound measurement. Explain the reason and name what evidence or approval would be necessary to revisit it.
## Review gate
The generated plan always starts `approval.status: pending`. A human reviewer can approve implementation outside this skill. Rerun validation after any plan or input-bundle change; validation confirms artifact integrity and traceability, not business approval or platform behavior.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |