Update to Omneky
Snapshot Oct 27, 2026 · 00:28 UTC · version 2.0.31
Collection source: downloaded plugin package.
Payment or plan references changed
Skill instructions — Now says: Use this when sign-in fails, the brand is unknown, a call returns an error state, or the user asks what they can do. One job: get the session working again...
Observed in published text. Live prices and checkout terms have not been verified by this change.
Up to 3 plugins free · no card required.
All fields and original values.
Skill instructions
Oct 24, 2026 · 13:12 UTC
`omneky-product-video`, `omneky-edit-and-resize`), launch, Handoff fields: [handoff-contract](../_shared/references/handoff-contract.md). Sourced ids (how to obtain / what to do if absent): [preconditions](../_shared/references/precondit...
Oct 27, 2026 · 00:28 UTC
Use this when sign-in fails, the brand is unknown, a call returns an error state, or the user asks what they can do. One job: get the session working again and point to the owning skill. `omneky-product-video`, `omneky-avatar-ugc`, `omne...
Supporting files
Oct 24, 2026 · 13:12 UTC
[]
Oct 27, 2026 · 00:28 UTC
[{"relative_path":"references/named-presets.md","size_in_bytes":4306},{"relative_path":"references/playbook.md","size_in_bytes":8974}]
Compare saved observations
Download comparison JSONSKILL.md text changes
− removed · + added
--- before +++ after @@ -18,6 +18,10 @@ # Session context and failure recovery +Use this when sign-in fails, the brand is unknown, a call returns an +error state, or the user asks what they can do. One job: get the +session working again and point to the owning skill. + <!-- host:begin name="oauth-sign-in" --> The user signs in to the connected Omneky MCP server through the host's OAuth. Never ask for or paste a JWT, token, or password. @@ -25,28 +29,21 @@ <!-- host:begin name="recovery-routing" --> This skill is the recovery entry. Healthy creative (`omneky-image-ads`, -`omneky-product-video`, `omneky-edit-and-resize`), launch, +`omneky-product-video`, `omneky-avatar-ugc`, +`omneky-voice-and-video-clone`, `omneky-edit-and-resize`), launch, analytics, catalogue, billing, and marketing-plan requests go straight to those skills. <!-- host:end --> -Handoff fields: -[handoff-contract](../_shared/references/handoff-contract.md). -Sourced ids (how to obtain / what to do if absent): -[preconditions](../_shared/references/preconditions.md). +Handoff fields: [handoff-contract](../_shared/references/handoff-contract.md). +Sourced ids (how to obtain / what to do if absent): [preconditions](../_shared/references/preconditions.md). <!-- host:begin name="scan-tools-routing" --> -Quote, retry, and polls: -[creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md). -Acceptance gates: -[creative-quality-gates](../_shared/references/creative-quality-gates.md). -Structured brief: -[creative-brief](../_shared/references/creative-brief.md). -QA and cheapest repair: -[media-qa-and-repair](../_shared/references/media-qa-and-repair.md). -Named presets and other intents: -[capability-registry](../_shared/references/capability-registry.md). -Write for: -[marketer-persona](../_shared/references/marketer-persona.md). +Quote, retry, and polls: [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md). +Acceptance gates: [creative-quality-gates](../_shared/references/creative-quality-gates.md). +Structured brief: [creative-brief](../_shared/references/creative-brief.md). +QA and cheapest repair: [media-qa-and-repair](../_shared/references/media-qa-and-repair.md). +Named presets and other intents: [capability-registry](../_shared/references/capability-registry.md). +Write for: [marketer-persona](../_shared/references/marketer-persona.md). <!-- host:end --> ## Goal @@ -59,14 +56,23 @@ | Need | Check | If missing | |---|---|---| -| Session | `get_current_user` | Host OAuth reconnect | +| Session | `get_account` | Host OAuth reconnect | | Brand | `list_brands` | Ask which; no brands → Omneky app | -| Connectors | `list_connector_statuses` | Connect via `omneky-launch-and-connect` | -| Credits (shortfall) | `get_credit_history` / `get_billing_summary` | `omneky-billing` (read only); add credits in the Omneky app | +| Connectors | `list_connections` | Connect via `omneky-launch-and-connect` | +| Credits (shortfall) | `get_account` | `omneky-billing` | ## Flow -Open the named-preset menu when they ask what they can do. Recover failures below. Show: what broke, the fix, what to do next. Do not quiz them. +1. Act on the request; there is no mandatory first-turn sequence. +2. Reuse the chosen `brand_id`. If unknown, `list_brands` (pass + `search`). No brands → Omneky app. Several matches → list name and + id and ask. One match → say it and keep it. Never switch brands + because metrics are empty. +3. On 401, ask the user to reconnect Omneky in the host. +4. Match the error to its recovery in + [playbook](references/playbook.md) (Failures; creative states + such as pending, terminal failure, and ambiguous paid submits). +5. Tell the user what broke, the fix, and the next step. ## When to use @@ -74,7 +80,6 @@ - Empty metrics, `status=timeout` / `pending` / `needs_user_decision` - `credit_insufficient` <!-- host:begin name="creative-when-to-use" --> -- `price_changed` - A video pick that returns `narrative_unresolved` <!-- host:end --> @@ -95,281 +100,29 @@ <!-- host:end --> - Not a license to invent purchase, card-PAN, or warehouse tools -<!-- host:begin name="named-presets" --> ## Named presets -When they ask what they can do, list these one-liners, then open -the owning skill from the capability registry. Do not run a -healthy creative, launch, plan, or billing request here. - -| Preset | One-liner | -|---|---| -| `winners-this-week` | This week's top ROAS ads (calendar week to date). | -| `weekly-performance-report` | Coverage, winners, and movers for the stated window. | -| `fatigue-refresh` | Find fatigued Meta ads; brief 3 replacement hypotheses. | -| `top-roas-variations` | N distinct stills from verified top-ROAS winners. | -| `budget-reallocation` | Read-only move recommendation; apply only after a yes. | -| `hero-still` | One product-hero still from the catalogue image. | -| `before-after-still` | Two-state still; both states from approved evidence. | -| `thumbnail-still` | One still for a placement thumbnail. | -| `product-video-story` | Pick a narrative card, then generate. | -| `faceless-product-video` | Product video with no talking head. | -| `generate-now-video` | Skip the card list only when they say so. | -| `paused-campaign-from-winners` | Launch winners `PAUSED` after one yes. | -| `competitor-analysis` | Cited positioning, angles, and gaps. | -| `ab-test-create` | One hypothesis, one variable, `PAUSED`, recorded plan. | -| `ab-test-measure` | z-test readout; refuse a winner when data is thin. | -| `weekly-marketing-plan` | Goal → analyze → create → revise → preview; host runs steps. | -| `brand-setup-from-website` | Analyse the site, import products, stay on that brand. | -| `first-ad-today` | URL → brief → quote → `PAUSED` launch. | -| `roas-cpa-drop` | Movers + breakdown + daily; what to pause or scale. | -| `waste-pruner` | Underperformers; pause only after a yes. | -| `weekly-cross-brand-report` | Per-client cards; no silent brand switch. | -| `hero-to-all-sizes` | One hero, then every remaining placement size. | -| `seasonal-promo-kit` | One promo brief across the placements they named. | -| `product-launch-video` | Launch / explainer product video from catalogue stills. | -| `audience-targeting` | Breakdowns + `search_ad_targeting`; apply only after a yes. | -| `target-cpa-budget` | Floor from `minimum_budget_for_objective` / `reach_estimate`. | -| `gsc-queries-to-ad-angles` | Search Console queries → cited ad angles. | -| `keyword-gaps` | Semrush / Ahrefs gaps, both sides fetched. | -| `domain-rating-backlinks` | Ahrefs rating + backlink overview (not Ad Library). | -| `hubspot-attribution` | Campaign/deal labels side by side; inferred, not attributed. | -| `abm-from-hubspot` | Cited firmographic lines into a creative brief. | -| `multi-brand-rollout` | One brand per yes; never silent-switch. | -| `credits-and-cost` | `quote_creative_credits`; image 5, product video 30. | -| `credit-history-audit` | Ledger via `get_credit_history`. | - -Capability gaps (subtitles burn-in, website builder, try-on, -motion-craft, video edit, narrator-on-existing, YouTube pack, -video ad-multiplier) live in the capability registry. Decline -those asks. Do not fake them. -<!-- host:end --> - -## Context, not onboarding - -There is no mandatory first-turn sequence. Act on the user's request and -resolve only what it needs. Reuse known identity and brand context. -Ask only for missing info. - -1. **Identity.** The server knows who is signed in. Call - `get_current_user` only when the user asks who they are signed in as. - On JWT / 401, ask the user to reconnect the Omneky app in the host. -2. **Brand.** Reuse the `brand_id` already chosen in this conversation. - If it is unknown and the request needs one, `list_brands` (pass - `search` when the user named the brand): - - **No brands.** Say this login has no brands yet. Do not invent a - `brand_id`. Point them at the Omneky app (Ads Central) to create - one. - - **Several matches.** List each matching name and id and ask which - one. Do not take the first row. - - **One clear match.** Say that name and id, then keep it. -3. Stick to the chosen `brand_id`. Empty metrics are not a reason to - switch brands. -4. `get_brand_details` when the next step needs logo, colors, or copy. - Basic record only: `get_brand`. -5. `list_connector_statuses` when the user asks what is connected. -6. `health_check` only when diagnosing a connection problem. It does - not return ads, credits, brands, or ROAS. - -## Approvals - -Ask once, concretely, before anything paid or live: what will happen, -for which brand, and that the server will report the cost. Reads, -lookups, and status checks need no approval. Destructive deletes keep -their own explicit confirmation (`omneky-launch-and-connect`). - -<!-- host:begin name="creative-failure-recovery" --> -## Hard rules shared with creative - -- After a narrative card pick, pass that `narrative_id`. A bad or - missing id is `narrative_unresolved`: re-fetch narratives and pick - again. Do not turn that error into Generate Now. -- Generate Now is only a true omit of both narrative fields, and only - when the user explicitly skips the pick (`omneky-product-video`). -- A hero spin or product animation is not a commercial. Do not route - it to `generate_product_video`. -- `environment_setting` is an HTTPS image URL only. -- `generate_image_ad` uses `ad_concept` (never `prompt`) and a - `gpt_ad_gen_id` (`<short-slug>-<uuid4>`). New id per new variant; - see Ambiguous paid-submit recovery for retries. `edit_image` uses - `images` (list of URL strings) and `prompt`. -- Generation poll bounds live in - [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md). - Plan poll bounds live in `omneky-marketing-plan`. Do not - silent-loop here. -- You decide whether to quote (`quote_creative_credits`) and whether - to ask first; if you show a price, use the exact quoted total. - Never guess. Pass - `items[].expected_credits` on the paid tool when that field is - listed; omit it if not. Do not pass `expected_credits` for a figure - you did not get from a quote. `price_changed` means re-quote — do - not retry with the old figure. - -## Quote unavailable - -Whole-call miss: `quote_creative_credits` missing, errors, -`quote_unavailable`, or every line is `not_quotable` -(`total_credits` omitted or null). The price is not available: -never treat a missing or 0 total as the price. You decide whether -to proceed; say the price was not available and report credits -actually charged afterward. Mixed batch (some lines quoted, some -`not_quotable`): show -the quoted totals and pass no `expected_credits` for the -`not_quotable` line. - -## Ambiguous paid-submit recovery - -This rule covers async paid submits (`generate_image_ad` and -`generate_product_video`). Sync edit and resize are charged on -each call. - -### Rejected before acceptance - -These responses happen before any charge. Fix the request and -resubmit only when the code is in this list: - -- `schema_invalid` -- `invalid_argument` -- `price_changed` (re-run `quote_creative_credits`, get fresh - approval; do **not** retry with the old `expected_credits`) -- `not_found_or_forbidden` -- `credit_insufficient` (nothing charged; add credits in the Omneky app - or stop — do not - retry until they have credits) -- `narrative_unresolved` (re-fetch and pick again; no job) -- `quote_unavailable` (price cannot be confirmed; do not invent one) - -Mint a new submit id only after one of those recognized pre-accept -rejections or the user's OK. - -### Credit outcome - -Read `credits_charged` when the tool result includes it. - -- `credits_charged: false` means this server did not deduct. Say that. - Do not tell the user you cannot tell whether credits were deducted. -- `submission: not_started` with `credits_charged: false` means no job - exists. Fix the request and resubmit only for a recognized - pre-accept rejection (the list above, or this pair). -- `submission: unconfirmed` with `credits_charged: false` means credits - were not charged, but a job might still have been queued. Check - `get_generation_status` with the returned `job_id` and `kind=video`. - If no job is found and generate_* tools for this job type appear - on this session's `tools/list`, it is safe to retry once. If they - do not, do not retry. Poll `get_generation_status` or - `list_ad_instance_items` when listed, and tell the user generation - isn't available on this connector. You will not be - double charged because credits are only deducted after a confirmed - `job_id`. -- `credits_charged: true` means the charge landed and `job_id` is the - poll id. -- If `credits_charged` is absent, keep the rule below. - -### Do not resubmit - -Split image and video recovery. Do not reuse "the same id" for both. - -- Image ads: a timeout, 5xx, or connection drop means do **not** - mint a new id. Reuse the same `gpt_ad_gen_id`. Check - `get_generation_status` with `kind=image` (bounds in - [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md)). - A repeat of an already-charged image id returns - `already_submitted` and the original job. -- Product video `submission: unconfirmed` (timeout / 5xx): there is - no caller submission key. Check `get_generation_status` with the - returned `job_id` and `kind=video`. If no job is found and - generate_* tools for this job type appear on this session's - `tools/list`, it is safe to retry once. If they do not, do not - retry. Poll `get_generation_status` or `list_ad_instance_items` - when listed, and tell the user generation isn't available on - this connector. Credits deduct only after a confirmed `job_id`. -- Product video `submission: not_started`: no job exists. Fix the - request and resubmit only for a recognized pre-accept rejection. -- `edit_image`, `resize_ad`, and `resize_image`: do **not** - auto-retry. Check `get_credit_history` or the returned asset, - or ask the user. A retry is a new charge. -- Any other timeout, 5xx, connection drop, or unrecognized error - means do **not** resubmit. The request may already have been - accepted. - -An image transport retry reuses `gpt_ad_gen_id`. Product video -has no caller submission key — poll the returned `job_id`, and -retry once only if no job is found and generate_* tools for -this job type are on `tools/list`. If they are not, do not -retry; tell the user generation isn't available on this -connector. A new -image variant uses a new `gpt_ad_gen_id`. Never mint a new -`gpt_ad_gen_id` just because a call timed out or the connection -dropped. - -## After a job or a failed check - -Three states. Poll bounds live in -[creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md). -Acceptance gates live in -[creative-quality-gates](../_shared/references/creative-quality-gates.md). -Never reuse a completed job id to repair it. - -**Pending.** Still running. Same generation id. Never resubmit. - -**Terminal failure.** Job failed. Explain. A new attempt is a new -paid generation unless the tool says refunded / not charged; the -host decides whether to quote or ask first. - -**Quality failure.** Job succeeded but a check failed. Show the -result. Offer a revision. A new generation is a new paid run; the host decides whether to ask. -<!-- host:end --> - -## Failures - -- Empty reporting → empty is missing data, not zero. Narrow the - window; do not switch brands to force a number. Performance depth: - `omneky-analytics-and-catalogue`. -- `get_recommendations` or unscoped `get_ad_groups` timeout → narrow - the query; do not retry the identical call this turn. -- Scrape `pending` → retry later, same `brand_id` + url. -- LinkedIn / TikTok 400 → location pre-call on the launch skill - (`search_ad_targeting` / `get_tiktok_location_ids`). X Ads channel - value is `twitter`. -- `needs_user_decision` → `request_user_decision`. Present the - questions in chat. Never dump `clarification_needed` JSON. Never - auto-retry. -- `credit_insufficient` → nothing started or was charged. Tell the user - the shortfall (`required` vs `balance`). Credits are added in the - Omneky app; offer that or stop. - Never collect a PAN. Never invent a credit purchase tool. -- `price_changed` → the live quote differed. Nothing started or was - charged. Re-run `quote_creative_credits` (host decides whether to ask again), and - never auto-retry with the old `expected_credits`. If that field is - not on `tools/list`, omit it. -- A tool named in these skills is not available in this session: do - not invent it. Say it is not available here; the user can do it in - the Omneky app (Ads Central). Do not offer another app or connector. +When they ask what they can do, show the one-liners in +[named-presets](references/named-presets.md), then open the owning +skill. Do not run a healthy request here. ## Failure and recovery Auth / 401 → `omneky-getting-started-and-failures` (host OAuth). -Empty reporting → empty is missing data, not zero; narrow the window. -`credit_insufficient` → nothing charged; the user can add credits in the Omneky app. -`needs_user_decision` → `request_user_decision`; never dump JSON. -Missing tool → say so; Omneky app. Do not invent it. +Empty reporting → `get_performance_summary` (`include_coverage=true`); empty is not zero. +`credit_insufficient` → nothing charged; state balance vs required +(`omneky-billing`). No payment links. +`needs_user_decision` → fill the missing detail and call again; ask in plain chat only if you cannot. Never dump JSON. +Missing tool → this session may run an older published app snapshot: look +the name up in [tool-name-fallbacks](../_shared/references/tool-name-fallbacks.md) +and call the older name if that one is on `tools/list`. If neither is +listed, say so; Omneky app. Do not invent it. Never ask for internal ids. Never silent-switch `brand_id`. <!-- host:begin name="capability-gaps-link" --> ## Gaps -This connector cannot do comment/social inbox, landing pages, -subtitle burn-in, email/SMS, organic scheduling, Meta Ad Library -competitor ads, virtual try-on, or website builder. Decline those. -See [capability-gaps](../_shared/references/capability-gaps.md). +Decline social inbox, landing pages, subtitle burn-in, email/SMS, +organic scheduling, Ad Library ads, try-on, and website builder +([capability-gaps](../_shared/references/capability-gaps.md)). <!-- host:end --> - -## Credits on this listing - -The host decides whether a quote or a yes is needed before a paid call; -there is no fixed confirm step. Each image ad costs 5 credits and each -multi-scene product video costs 30 credits, charged when the job is -submitted. `quote_creative_credits` gives an exact number when that -helps (edits, resizes, batches). A `credit_insufficient` result means -the job did not start: tell the user to add credits in the Omneky app, -and do not retry until they do.
Full technical diff · 2 changed fields
changed /included_files
[]
[
{
"relative_path": "references/named-presets.md",
"size_in_bytes": 4306
},
{
"relative_path": "references/playbook.md",
"size_in_bytes": 8974
}
]changed /skill_md_contents
"---\nname: omneky-getting-started-and-failures\ndescription: >-\n Recover Omneky session context: sign-in/OAuth, unknown brand,\n unauthorized, pending/timeout/needs_user_decision/credit_insufficient.\n Also the named-preset menu when they ask what they can do. Do not\n use for first-run onboarding (omneky-onboarding), agency\n cross-brand (omneky-agency), weekly marketing plans\n (omneky-marketing-plan), paid-media analysis or catalogue\n (omneky-analytics-and-catalogue), competitor analysis\n (omneky-competitor-analysis), A/B tests (omneky-ab-tests),\n generating creatives, launch/connect\n (omneky-launch-and-connect), or credits (omneky-billing).\n Keywords: sign in, OAuth, list brands, unauthorized, timeout,\n empty data, no brands, which brand, needs user decision,\n credit insufficient, named presets, what can you do.\n---\n\n# Session context and failure recovery\n\n<!-- host:begin name=\"oauth-sign-in\" -->\nThe user signs in to the connected Omneky MCP server through the host's\nOAuth. Never ask for or paste a JWT, token, or password.\n<!-- host:end -->\n\n<!-- host:begin name=\"recovery-routing\" -->\nThis skill is the recovery entry. Healthy creative (`omneky-image-ads`,\n`omneky-product-video`, `omneky-edit-and-resize`), launch,\nanalytics, catalogue, billing, and marketing-plan requests go straight\nto those skills.\n<!-- host:end -->\n\nHandoff fields:\n[handoff-contract](../_shared/references/handoff-contract.md).\nSourced ids (how to obtain / what to do if absent):\n[preconditions](../_shared/references/preconditions.md).\n<!-- host:begin name=\"scan-tools-routing\" -->\nQuote, retry, and polls:\n[creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md).\nAcceptance gates:\n[creative-quality-gates](../_shared/references/creative-quality-gates.md).\nStructured brief:\n[creative-brief](../_shared/references/creative-brief.md).\nQA and cheapest repair:\n[media-qa-and-repair](../_shared/references/media-qa-and-repair.md).\nNamed presets and other intents:\n[capability-registry](../_shared/references/capability-registry.md).\nWrite for:\n[marketer-persona](../_shared/references/marketer-persona.md).\n<!-- host:end -->\n\n## Goal\n\n<!-- host:begin name=\"goal-first-run\" -->\nRecover a broken session and list named presets. First-run setup is `omneky-onboarding`.\n<!-- host:end -->\n\n## Prerequisites\n\n| Need | Check | If missing |\n|---|---|---|\n| Session | `get_current_user` | Host OAuth reconnect |\n| Brand | `list_brands` | Ask which; no brands → Omneky app |\n| Connectors | `list_connector_statuses` | Connect via `omneky-launch-and-connect` |\n| Credits (shortfall) | `get_credit_history` / `get_billing_summary` | `omneky-billing` (read only); add credits in the Omneky app |\n\n## Flow\n\nOpen the named-preset menu when they ask what they can do. Recover failures below. Show: what broke, the fix, what to do next. Do not quiz them.\n\n## When to use\n\n- 401 / JWT errors, unknown or ambiguous `brand_id`\n- Empty metrics, `status=timeout` / `pending` / `needs_user_decision`\n- `credit_insufficient`\n<!-- host:begin name=\"creative-when-to-use\" -->\n- `price_changed`\n- A video pick that returns `narrative_unresolved`\n<!-- host:end -->\n\n## When not to use\n\n<!-- host:begin name=\"creative-when-not\" -->\n- A healthy \"make a branded ad\" / \"video ad\" / \"which ads should I make\n more of\" / \"how many credits\" / \"weekly marketing plan\" request —\n open that skill directly\n<!-- host:end -->\n- Weekly plan create / revise / poll / approve →\n `omneky-marketing-plan` (this skill does not own that loop)\n- Competitor analysis → `omneky-competitor-analysis`\n<!-- host:begin name=\"omitted-host-skills\" -->\n- A/B test create or measure → `omneky-ab-tests`\n- First-run setup / first ad today → `omneky-onboarding`\n- Cross-brand / all clients → `omneky-agency`\n<!-- host:end -->\n- Not a license to invent purchase, card-PAN, or warehouse tools\n\n<!-- host:begin name=\"named-presets\" -->\n## Named presets\n\nWhen they ask what they can do, list these one-liners, then open\nthe owning skill from the capability registry. Do not run a\nhealthy creative, launch, plan, or billing request here.\n\n| Preset | One-liner |\n|---|---|\n| `winners-this-week` | This week's top ROAS ads (calendar week to date). |\n| `weekly-performance-report` | Coverage, winners, and movers for the stated window. |\n| `fatigue-refresh` | Find fatigued Meta ads; brief 3 replacement hypotheses. |\n| `top-roas-variations` | N distinct stills from verified top-ROAS winners. |\n| `budget-reallocation` | Read-only move recommendation; apply only after a yes. |\n| `hero-still` | One product-hero still from the catalogue image. |\n| `before-after-still` | Two-state still; both states from approved evidence. |\n| `thumbnail-still` | One still for a placement thumbnail. |\n| `product-video-story` | Pick a narrative card, then generate. |\n| `faceless-product-video` | Product video with no talking head. |\n| `generate-now-video` | Skip the card list only when they say so. |\n| `paused-campaign-from-winners` | Launch winners `PAUSED` after one yes. |\n| `competitor-analysis` | Cited positioning, angles, and gaps. |\n| `ab-test-create` | One hypothesis, one variable, `PAUSED`, recorded plan. |\n| `ab-test-measure` | z-test readout; refuse a winner when data is thin. |\n| `weekly-marketing-plan` | Goal → analyze → create → revise → preview; host runs steps. |\n| `brand-setup-from-website` | Analyse the site, import products, stay on that brand. |\n| `first-ad-today` | URL → brief → quote → `PAUSED` launch. |\n| `roas-cpa-drop` | Movers + breakdown + daily; what to pause or scale. |\n| `waste-pruner` | Underperformers; pause only after a yes. |\n| `weekly-cross-brand-report` | Per-client cards; no silent brand switch. |\n| `hero-to-all-sizes` | One hero, then every remaining placement size. |\n| `seasonal-promo-kit` | One promo brief across the placements they named. |\n| `product-launch-video` | Launch / explainer product video from catalogue stills. |\n| `audience-targeting` | Breakdowns + `search_ad_targeting`; apply only after a yes. |\n| `target-cpa-budget` | Floor from `minimum_budget_for_objective` / `reach_estimate`. |\n| `gsc-queries-to-ad-angles` | Search Console queries → cited ad angles. |\n| `keyword-gaps` | Semrush / Ahrefs gaps, both sides fetched. |\n| `domain-rating-backlinks` | Ahrefs rating + backlink overview (not Ad Library). |\n| `hubspot-attribution` | Campaign/deal labels side by side; inferred, not attributed. |\n| `abm-from-hubspot` | Cited firmographic lines into a creative brief. |\n| `multi-brand-rollout` | One brand per yes; never silent-switch. |\n| `credits-and-cost` | `quote_creative_credits`; image 5, product video 30. |\n| `credit-history-audit` | Ledger via `get_credit_history`. |\n\nCapability gaps (subtitles burn-in, website builder, try-on,\nmotion-craft, video edit, narrator-on-existing, YouTube pack,\nvideo ad-multiplier) live in the capability registry. Decline\nthose asks. Do not fake them.\n<!-- host:end -->\n\n## Context, not onboarding\n\nThere is no mandatory first-turn sequence. Act on the user's request and\nresolve only what it needs. Reuse known identity and brand context.\nAsk only for missing info.\n\n1. **Identity.** The server knows who is signed in. Call\n `get_current_user` only when the user asks who they are signed in as.\n On JWT / 401, ask the user to reconnect the Omneky app in the host.\n2. **Brand.** Reuse the `brand_id` already chosen in this conversation.\n If it is unknown and the request needs one, `list_brands` (pass\n `search` when the user named the brand):\n - **No brands.** Say this login has no brands yet. Do not invent a\n `brand_id`. Point them at the Omneky app (Ads Central) to create\n one.\n - **Several matches.** List each matching name and id and ask which\n one. Do not take the first row.\n - **One clear match.** Say that name and id, then keep it.\n3. Stick to the chosen `brand_id`. Empty metrics are not a reason to\n switch brands.\n4. `get_brand_details` when the next step needs logo, colors, or copy.\n Basic record only: `get_brand`.\n5. `list_connector_statuses` when the user asks what is connected.\n6. `health_check` only when diagnosing a connection problem. It does\n not return ads, credits, brands, or ROAS.\n\n## Approvals\n\nAsk once, concretely, before anything paid or live: what will happen,\nfor which brand, and that the server will report the cost. Reads,\nlookups, and status checks need no approval. Destructive deletes keep\ntheir own explicit confirmation (`omneky-launch-and-connect`).\n\n<!-- host:begin name=\"creative-failure-recovery\" -->\n## Hard rules shared with creative\n\n- After a narrative card pick, pass that `narrative_id`. A bad or\n missing id is `narrative_unresolved`: re-fetch narratives and pick\n again. Do not turn that error into Generate Now.\n- Generate Now is only a true omit of both narrative fields, and only\n when the user explicitly skips the pick (`omneky-product-video`).\n- A hero spin or product animation is not a commercial. Do not route\n it to `generate_product_video`.\n- `environment_setting` is an HTTPS image URL only.\n- `generate_image_ad` uses `ad_concept` (never `prompt`) and a\n `gpt_ad_gen_id` (`<short-slug>-<uuid4>`). New id per new variant;\n see Ambiguous paid-submit recovery for retries. `edit_image` uses\n `images` (list of URL strings) and `prompt`.\n- Generation poll bounds live in\n [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md).\n Plan poll bounds live in `omneky-marketing-plan`. Do not\n silent-loop here.\n- You decide whether to quote (`quote_creative_credits`) and whether\n to ask first; if you show a price, use the exact quoted total.\n Never guess. Pass\n `items[].expected_credits` on the paid tool when that field is\n listed; omit it if not. Do not pass `expected_credits` for a figure\n you did not get from a quote. `price_changed` means re-quote — do\n not retry with the old figure.\n\n## Quote unavailable\n\nWhole-call miss: `quote_creative_credits` missing, errors,\n`quote_unavailable`, or every line is `not_quotable`\n(`total_credits` omitted or null). The price is not available:\nnever treat a missing or 0 total as the price. You decide whether\nto proceed; say the price was not available and report credits\nactually charged afterward. Mixed batch (some lines quoted, some\n`not_quotable`): show\nthe quoted totals and pass no `expected_credits` for the\n`not_quotable` line.\n\n## Ambiguous paid-submit recovery\n\nThis rule covers async paid submits (`generate_image_ad` and\n`generate_product_video`). Sync edit and resize are charged on\neach call.\n\n### Rejected before acceptance\n\nThese responses happen before any charge. Fix the request and\nresubmit only when the code is in this list:\n\n- `schema_invalid`\n- `invalid_argument`\n- `price_changed` (re-run `quote_creative_credits`, get fresh\n approval; do **not** retry with the old `expected_credits`)\n- `not_found_or_forbidden`\n- `credit_insufficient` (nothing charged; add credits in the Omneky app\n or stop — do not\n retry until they have credits)\n- `narrative_unresolved` (re-fetch and pick again; no job)\n- `quote_unavailable` (price cannot be confirmed; do not invent one)\n\nMint a new submit id only after one of those recognized pre-accept\nrejections or the user's OK.\n\n### Credit outcome\n\nRead `credits_charged` when the tool result includes it.\n\n- `credits_charged: false` means this server did not deduct. Say that.\n Do not tell the user you cannot tell whether credits were deducted.\n- `submission: not_started` with `credits_charged: false` means no job\n exists. Fix the request and resubmit only for a recognized\n pre-accept rejection (the list above, or this pair).\n- `submission: unconfirmed` with `credits_charged: false` means credits\n were not charged, but a job might still have been queued. Check\n `get_generation_status` with the returned `job_id` and `kind=video`.\n If no job is found and generate_* tools for this job type appear\n on this session's `tools/list`, it is safe to retry once. If they\n do not, do not retry. Poll `get_generation_status` or\n `list_ad_instance_items` when listed, and tell the user generation\n isn't available on this connector. You will not be\n double charged because credits are only deducted after a confirmed\n `job_id`.\n- `credits_charged: true` means the charge landed and `job_id` is the\n poll id.\n- If `credits_charged` is absent, keep the rule below.\n\n### Do not resubmit\n\nSplit image and video recovery. Do not reuse \"the same id\" for both.\n\n- Image ads: a timeout, 5xx, or connection drop means do **not**\n mint a new id. Reuse the same `gpt_ad_gen_id`. Check\n `get_generation_status` with `kind=image` (bounds in\n [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md)).\n A repeat of an already-charged image id returns\n `already_submitted` and the original job.\n- Product video `submission: unconfirmed` (timeout / 5xx): there is\n no caller submission key. Check `get_generation_status` with the\n returned `job_id` and `kind=video`. If no job is found and\n generate_* tools for this job type appear on this session's\n `tools/list`, it is safe to retry once. If they do not, do not\n retry. Poll `get_generation_status` or `list_ad_instance_items`\n when listed, and tell the user generation isn't available on\n this connector. Credits deduct only after a confirmed `job_id`.\n- Product video `submission: not_started`: no job exists. Fix the\n request and resubmit only for a recognized pre-accept rejection.\n- `edit_image`, `resize_ad`, and `resize_image`: do **not**\n auto-retry. Check `get_credit_history` or the returned asset,\n or ask the user. A retry is a new charge.\n- Any other timeout, 5xx, connection drop, or unrecognized error\n means do **not** resubmit. The request may already have been\n accepted.\n\nAn image transport retry reuses `gpt_ad_gen_id`. Product video\nhas no caller submission key — poll the returned `job_id`, and\nretry once only if no job is found and generate_* tools for\nthis job type are on `tools/list`. If they are not, do not\nretry; tell the user generation isn't available on this\nconnector. A new\nimage variant uses a new `gpt_ad_gen_id`. Never mint a new\n`gpt_ad_gen_id` just because a call timed out or the connection\ndropped.\n\n## After a job or a failed check\n\nThree states. Poll bounds live in\n[creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md).\nAcceptance gates live in\n[creative-quality-gates](../_shared/references/creative-quality-gates.md).\nNever reuse a completed job id to repair it.\n\n**Pending.** Still running. Same generation id. Never resubmit.\n\n**Terminal failure.** Job failed. Explain. A new attempt is a new\npaid generation unless the tool says refunded / not charged; the\nhost decides whether to quote or ask first.\n\n**Quality failure.** Job succeeded but a check failed. Show the\nresult. Offer a revision. A new generation is a new paid run; the host decides whether to ask.\n<!-- host:end -->\n\n## Failures\n\n- Empty reporting → empty is missing data, not zero. Narrow the\n window; do not switch brands to force a number. Performance depth:\n `omneky-analytics-and-catalogue`.\n- `get_recommendations` or unscoped `get_ad_groups` timeout → narrow\n the query; do not retry the identical call this turn.\n- Scrape `pending` → retry later, same `brand_id` + url.\n- LinkedIn / TikTok 400 → location pre-call on the launch skill\n (`search_ad_targeting` / `get_tiktok_location_ids`). X Ads channel\n value is `twitter`.\n- `needs_user_decision` → `request_user_decision`. Present the\n questions in chat. Never dump `clarification_needed` JSON. Never\n auto-retry.\n- `credit_insufficient` → nothing started or was charged. Tell the user\n the shortfall (`required` vs `balance`). Credits are added in the\n Omneky app; offer that or stop.\n Never collect a PAN. Never invent a credit purchase tool.\n- `price_changed` → the live quote differed. Nothing started or was\n charged. Re-run `quote_creative_credits` (host decides whether to ask again), and\n never auto-retry with the old `expected_credits`. If that field is\n not on `tools/list`, omit it.\n- A tool named in these skills is not available in this session: do\n not invent it. Say it is not available here; the user can do it in\n the Omneky app (Ads Central). Do not offer another app or connector.\n\n## Failure and recovery\n\nAuth / 401 → `omneky-getting-started-and-failures` (host OAuth).\nEmpty reporting → empty is missing data, not zero; narrow the window.\n`credit_insufficient` → nothing charged; the user can add credits in the Omneky app.\n`needs_user_decision` → `request_user_decision`; never dump JSON.\nMissing tool → say so; Omneky app. Do not invent it.\nNever ask for internal ids. Never silent-switch `brand_id`.\n\n<!-- host:begin name=\"capability-gaps-link\" -->\n## Gaps\n\nThis connector cannot do comment/social inbox, landing pages,\nsubtitle burn-in, email/SMS, organic scheduling, Meta Ad Library\ncompetitor ads, virtual try-on, or website builder. Decline those.\nSee [capability-gaps](../_shared/references/capability-gaps.md).\n<!-- host:end -->\n\n## Credits on this listing\n\nThe host decides whether a quote or a yes is needed before a paid call;\nthere is no fixed confirm step. Each image ad costs 5 credits and each\nmulti-scene product video costs 30 credits, charged when the job is\nsubmitted. `quote_creative_credits` gives an exact number when that\nhelps (edits, resizes, batches). A `credit_insufficient` result means\nthe job did not start: tell the user to add credits in the Omneky app,\nand do not retry until they do.\n"
"---\nname: omneky-getting-started-and-failures\ndescription: >-\n Recover Omneky session context: sign-in/OAuth, unknown brand,\n unauthorized, pending/timeout/needs_user_decision/credit_insufficient.\n Also the named-preset menu when they ask what they can do. Do not\n use for first-run onboarding (omneky-onboarding), agency\n cross-brand (omneky-agency), weekly marketing plans\n (omneky-marketing-plan), paid-media analysis or catalogue\n (omneky-analytics-and-catalogue), competitor analysis\n (omneky-competitor-analysis), A/B tests (omneky-ab-tests),\n generating creatives, launch/connect\n (omneky-launch-and-connect), or credits (omneky-billing).\n Keywords: sign in, OAuth, list brands, unauthorized, timeout,\n empty data, no brands, which brand, needs user decision,\n credit insufficient, named presets, what can you do.\n---\n\n# Session context and failure recovery\n\nUse this when sign-in fails, the brand is unknown, a call returns an\nerror state, or the user asks what they can do. One job: get the\nsession working again and point to the owning skill.\n\n<!-- host:begin name=\"oauth-sign-in\" -->\nThe user signs in to the connected Omneky MCP server through the host's\nOAuth. Never ask for or paste a JWT, token, or password.\n<!-- host:end -->\n\n<!-- host:begin name=\"recovery-routing\" -->\nThis skill is the recovery entry. Healthy creative (`omneky-image-ads`,\n`omneky-product-video`, `omneky-avatar-ugc`,\n`omneky-voice-and-video-clone`, `omneky-edit-and-resize`), launch,\nanalytics, catalogue, billing, and marketing-plan requests go straight\nto those skills.\n<!-- host:end -->\n\nHandoff fields: [handoff-contract](../_shared/references/handoff-contract.md).\nSourced ids (how to obtain / what to do if absent): [preconditions](../_shared/references/preconditions.md).\n<!-- host:begin name=\"scan-tools-routing\" -->\nQuote, retry, and polls: [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md).\nAcceptance gates: [creative-quality-gates](../_shared/references/creative-quality-gates.md).\nStructured brief: [creative-brief](../_shared/references/creative-brief.md).\nQA and cheapest repair: [media-qa-and-repair](../_shared/references/media-qa-and-repair.md).\nNamed presets and other intents: [capability-registry](../_shared/references/capability-registry.md).\nWrite for: [marketer-persona](../_shared/references/marketer-persona.md).\n<!-- host:end -->\n\n## Goal\n\n<!-- host:begin name=\"goal-first-run\" -->\nRecover a broken session and list named presets. First-run setup is `omneky-onboarding`.\n<!-- host:end -->\n\n## Prerequisites\n\n| Need | Check | If missing |\n|---|---|---|\n| Session | `get_account` | Host OAuth reconnect |\n| Brand | `list_brands` | Ask which; no brands → Omneky app |\n| Connectors | `list_connections` | Connect via `omneky-launch-and-connect` |\n| Credits (shortfall) | `get_account` | `omneky-billing` |\n\n## Flow\n\n1. Act on the request; there is no mandatory first-turn sequence.\n2. Reuse the chosen `brand_id`. If unknown, `list_brands` (pass\n `search`). No brands → Omneky app. Several matches → list name and\n id and ask. One match → say it and keep it. Never switch brands\n because metrics are empty.\n3. On 401, ask the user to reconnect Omneky in the host.\n4. Match the error to its recovery in\n [playbook](references/playbook.md) (Failures; creative states\n such as pending, terminal failure, and ambiguous paid submits).\n5. Tell the user what broke, the fix, and the next step.\n\n## When to use\n\n- 401 / JWT errors, unknown or ambiguous `brand_id`\n- Empty metrics, `status=timeout` / `pending` / `needs_user_decision`\n- `credit_insufficient`\n<!-- host:begin name=\"creative-when-to-use\" -->\n- A video pick that returns `narrative_unresolved`\n<!-- host:end -->\n\n## When not to use\n\n<!-- host:begin name=\"creative-when-not\" -->\n- A healthy \"make a branded ad\" / \"video ad\" / \"which ads should I make\n more of\" / \"how many credits\" / \"weekly marketing plan\" request —\n open that skill directly\n<!-- host:end -->\n- Weekly plan create / revise / poll / approve →\n `omneky-marketing-plan` (this skill does not own that loop)\n- Competitor analysis → `omneky-competitor-analysis`\n<!-- host:begin name=\"omitted-host-skills\" -->\n- A/B test create or measure → `omneky-ab-tests`\n- First-run setup / first ad today → `omneky-onboarding`\n- Cross-brand / all clients → `omneky-agency`\n<!-- host:end -->\n- Not a license to invent purchase, card-PAN, or warehouse tools\n\n## Named presets\n\nWhen they ask what they can do, show the one-liners in\n[named-presets](references/named-presets.md), then open the owning\nskill. Do not run a healthy request here.\n\n## Failure and recovery\n\nAuth / 401 → `omneky-getting-started-and-failures` (host OAuth).\nEmpty reporting → `get_performance_summary` (`include_coverage=true`); empty is not zero.\n`credit_insufficient` → nothing charged; state balance vs required\n(`omneky-billing`). No payment links.\n`needs_user_decision` → fill the missing detail and call again; ask in plain chat only if you cannot. Never dump JSON.\nMissing tool → this session may run an older published app snapshot: look\nthe name up in [tool-name-fallbacks](../_shared/references/tool-name-fallbacks.md)\nand call the older name if that one is on `tools/list`. If neither is\nlisted, say so; Omneky app. Do not invent it.\nNever ask for internal ids. Never silent-switch `brand_id`.\n\n<!-- host:begin name=\"capability-gaps-link\" -->\n## Gaps\n\nDecline social inbox, landing pages, subtitle burn-in, email/SMS,\norganic scheduling, Ad Library ads, try-on, and website builder\n([capability-gaps](../_shared/references/capability-gaps.md)).\n<!-- host:end -->\n"
Full snapshot data
{
"description": "Recover Omneky session context: sign-in/OAuth, unknown brand, unauthorized, pending/timeout/needs_user_decision/credit_insufficient. Also the named-preset menu when they ask what they can do. Do not use for first-run onboarding (omneky-onboarding), agency cross-brand (omneky-agency), weekly marketing plans (omneky-marketing-plan), paid-media analysis or catalogue (omneky-analytics-and-catalogue), competitor analysis (omneky-competitor-analysis), A/B tests (omneky-ab-tests), generating creatives, launch/connect (omneky-launch-and-connect), or credits (omneky-billing). Keywords: sign in, OAuth, list brands, unauthorized, timeout, empty data, no brands, which brand, needs user decision, credit insufficient, named presets, what can you do.",
"included_files": [
{
"relative_path": "references/named-presets.md",
"size_in_bytes": 4306
},
{
"relative_path": "references/playbook.md",
"size_in_bytes": 8974
}
],
"name": "omneky-getting-started-and-failures",
"skill_md_contents": "---\nname: omneky-getting-started-and-failures\ndescription: >-\n Recover Omneky session context: sign-in/OAuth, unknown brand,\n unauthorized, pending/timeout/needs_user_decision/credit_insufficient.\n Also the named-preset menu when they ask what they can do. Do not\n use for first-run onboarding (omneky-onboarding), agency\n cross-brand (omneky-agency), weekly marketing plans\n (omneky-marketing-plan), paid-media analysis or catalogue\n (omneky-analytics-and-catalogue), competitor analysis\n (omneky-competitor-analysis), A/B tests (omneky-ab-tests),\n generating creatives, launch/connect\n (omneky-launch-and-connect), or credits (omneky-billing).\n Keywords: sign in, OAuth, list brands, unauthorized, timeout,\n empty data, no brands, which brand, needs user decision,\n credit insufficient, named presets, what can you do.\n---\n\n# Session context and failure recovery\n\nUse this when sign-in fails, the brand is unknown, a call returns an\nerror state, or the user asks what they can do. One job: get the\nsession working again and point to the owning skill.\n\n<!-- host:begin name=\"oauth-sign-in\" -->\nThe user signs in to the connected Omneky MCP server through the host's\nOAuth. Never ask for or paste a JWT, token, or password.\n<!-- host:end -->\n\n<!-- host:begin name=\"recovery-routing\" -->\nThis skill is the recovery entry. Healthy creative (`omneky-image-ads`,\n`omneky-product-video`, `omneky-avatar-ugc`,\n`omneky-voice-and-video-clone`, `omneky-edit-and-resize`), launch,\nanalytics, catalogue, billing, and marketing-plan requests go straight\nto those skills.\n<!-- host:end -->\n\nHandoff fields: [handoff-contract](../_shared/references/handoff-contract.md).\nSourced ids (how to obtain / what to do if absent): [preconditions](../_shared/references/preconditions.md).\n<!-- host:begin name=\"scan-tools-routing\" -->\nQuote, retry, and polls: [creative-pricing-and-polling](../_shared/references/creative-pricing-and-polling.md).\nAcceptance gates: [creative-quality-gates](../_shared/references/creative-quality-gates.md).\nStructured brief: [creative-brief](../_shared/references/creative-brief.md).\nQA and cheapest repair: [media-qa-and-repair](../_shared/references/media-qa-and-repair.md).\nNamed presets and other intents: [capability-registry](../_shared/references/capability-registry.md).\nWrite for: [marketer-persona](../_shared/references/marketer-persona.md).\n<!-- host:end -->\n\n## Goal\n\n<!-- host:begin name=\"goal-first-run\" -->\nRecover a broken session and list named presets. First-run setup is `omneky-onboarding`.\n<!-- host:end -->\n\n## Prerequisites\n\n| Need | Check | If missing |\n|---|---|---|\n| Session | `get_account` | Host OAuth reconnect |\n| Brand | `list_brands` | Ask which; no brands → Omneky app |\n| Connectors | `list_connections` | Connect via `omneky-launch-and-connect` |\n| Credits (shortfall) | `get_account` | `omneky-billing` |\n\n## Flow\n\n1. Act on the request; there is no mandatory first-turn sequence.\n2. Reuse the chosen `brand_id`. If unknown, `list_brands` (pass\n `search`). No brands → Omneky app. Several matches → list name and\n id and ask. One match → say it and keep it. Never switch brands\n because metrics are empty.\n3. On 401, ask the user to reconnect Omneky in the host.\n4. Match the error to its recovery in\n [playbook](references/playbook.md) (Failures; creative states\n such as pending, terminal failure, and ambiguous paid submits).\n5. Tell the user what broke, the fix, and the next step.\n\n## When to use\n\n- 401 / JWT errors, unknown or ambiguous `brand_id`\n- Empty metrics, `status=timeout` / `pending` / `needs_user_decision`\n- `credit_insufficient`\n<!-- host:begin name=\"creative-when-to-use\" -->\n- A video pick that returns `narrative_unresolved`\n<!-- host:end -->\n\n## When not to use\n\n<!-- host:begin name=\"creative-when-not\" -->\n- A healthy \"make a branded ad\" / \"video ad\" / \"which ads should I make\n more of\" / \"how many credits\" / \"weekly marketing plan\" request —\n open that skill directly\n<!-- host:end -->\n- Weekly plan create / revise / poll / approve →\n `omneky-marketing-plan` (this skill does not own that loop)\n- Competitor analysis → `omneky-competitor-analysis`\n<!-- host:begin name=\"omitted-host-skills\" -->\n- A/B test create or measure → `omneky-ab-tests`\n- First-run setup / first ad today → `omneky-onboarding`\n- Cross-brand / all clients → `omneky-agency`\n<!-- host:end -->\n- Not a license to invent purchase, card-PAN, or warehouse tools\n\n## Named presets\n\nWhen they ask what they can do, show the one-liners in\n[named-presets](references/named-presets.md), then open the owning\nskill. Do not run a healthy request here.\n\n## Failure and recovery\n\nAuth / 401 → `omneky-getting-started-and-failures` (host OAuth).\nEmpty reporting → `get_performance_summary` (`include_coverage=true`); empty is not zero.\n`credit_insufficient` → nothing charged; state balance vs required\n(`omneky-billing`). No payment links.\n`needs_user_decision` → fill the missing detail and call again; ask in plain chat only if you cannot. Never dump JSON.\nMissing tool → this session may run an older published app snapshot: look\nthe name up in [tool-name-fallbacks](../_shared/references/tool-name-fallbacks.md)\nand call the older name if that one is on `tools/list`. If neither is\nlisted, say so; Omneky app. Do not invent it.\nNever ask for internal ids. Never silent-switch `brand_id`.\n\n<!-- host:begin name=\"capability-gaps-link\" -->\n## Gaps\n\nDecline social inbox, landing pages, subtitle burn-in, email/SMS,\norganic scheduling, Ad Library ads, try-on, and website builder\n([capability-gaps](../_shared/references/capability-gaps.md)).\n<!-- host:end -->\n"
}SHA-256 of public snapshot: 9b080e826c1a9b16ab632383ec065560a4df017afe852419be42bc29dd66255b