gooseworks 0.5.9-dev.58.1cb81d21 → 0.5.9-dev.59.2b371ca1

package/dist/skills/server-entry-skills.d.ts

@@ -1,8 +1,8 @@
 /** The rulebook version this text was built from. */
 export declare const SERVER_RULEBOOK_VERSION = "1.3.2";
 export declare const SERVER_VIDEO_ENTRY_SKILLS: {
-    readonly 'goose-video': "---\nname: goose-video\ndescription: \"Make a video ad in this chat. Study a supplied video reference before comparing styles; follow the next_step in every answer.\"\nversion: 4.4.4\n---\n\n# GooseWorks video\n\nFrom an installed entry, first catalog_fetch type skill goose-video on this same connection and follow that body. An entry already fetched in this run needs no repeat fetch.\n\nFor a video already started, call video_read and follow its next_step.\n\nExplicit custom: skip steps 5–6. Keep their mechanics and assets; research the brand/product and study the reference for its custom plan. Check current custom support before any write.\n\nFor other new videos start with get_started (wants video, request). A card is not a style choice or proof of viewing. Follow the reference rule before style selection or a plan.\n\n1. The request. If they haven't said what the video is about, ask in one short question.\n2. The brand. Ask which brand only when brand_read lists several; with one, use it.\n3. The product. Named or none wanted: find its id with brand_read (sections: products), or use []. Otherwise the product card asks: say one line and wait. Without cards, ask which product by name; use the only one, or go on with none. Never ask again after a pick.\n4. Research. Read the brand kit, learnings and the chosen products with brand_read (sections kit and learnings, plus the products' ids), then knowledge_search for them. Note what to say and avoid.\n5. The styles. Call video_formats with the request and the products' ids ([] for none). The customer picks; never recommend one. A style they named that the answer lists as available is their pick. Otherwise, with the style card on screen, say one short line and wait; without one, print the style table from the answer as it is and ask nothing else.\n6. The plan. Call video_create with the picked style and the products; its next_step says what follows.\n\n## A supplied video reference comes first\n\nWatch the actual video before choosing a style or writing a custom plan. Play through it, or inspect timestamped frames spanning it plus its audio/transcript. A URL, title, caption or thumbnail is not a viewing. Record the hook, scene order, pacing, narration/dialogue, framing, product interactions, text and ending, with timestamps and gaps. If access or sampling cannot answer an essential detail, ask for an accessible video or their choice to continue from the brief; never claim it was watched.\nWhen they have not chosen custom, call video_formats after brand research. For plausible matches, catalog_fetch type template returns the style's plan_note, scenes, duration, cast, sound and assets. Compare these against the observed reference and their essential choices. A similar title or high brand ranking does not prove a match. Explain fit or mismatch before they choose. Keep an explicit style choice; do not quietly change constraints or over-adapt it. Reassess a later reference before creating or rewriting a plan.\nIf no available style fits, or they explicitly want custom, catalog_fetch type skill make-custom-video on this connection before a project write. Require its complete production-harness dependency and current custom schemas. Follow its separate authenticated script and ingredient approvals, cumulative budgets and verified execution capability. If anything is unavailable, say so and keep the brief; do not substitute a template, import the reference as a finished video, use retired actions or switch environments. Fetching a skill alone proves no execution capability.\n\n## Rules\n\n- Talk to marketers: what they get and decide, never how it's made.\n- Never mention tools, ids, JSON, commands, software, vendors, models, timings, retries or your checks.\n- Everything happens here: never ask the customer to install anything or use a CLI, slash command or other app, except a video's make command when a next_step gives it (its last line says who starts it; original videos follow their own). Never give any other command, sign-in step or setting, and never ask them a technical question.\n- Money is a credit number exactly as an answer gives it: never dollars or your own math.\n- Follow the next_step in every answer.\n- Card on screen (display_hint \"widget\"): one short line at most, then the wait in its next_step or end your turn; never re-list it or open a question box.\n- Paid work needs the customer's yes to the credit total first, on the card or in chat at the price shown; never decide it for them, and pass on a yes only once they gave it to that total.\n- A yes is to the credit total the customer was shown: their press on the plan card, or their yes in chat sent with that total.\n- Where the app shows no card, print the plan and its price and ask for the customer's yes here; never send them to a website to approve.\n- Never invent an offer, a claim or proof.\n- GooseWorks holds every provider key and bills the work: never ask for a key, token, password or ad account id, and never call a vendor yourself.\n";
-    readonly 'make-custom-video': "---\nname: make-custom-video\ndescription: \"Make a custom video from its brief and studied reference, with authenticated script and ingredient reviews.\"\nversion: 3.5.1\nrequires_skills: [video-production-harness]\nharness_binding: gooseworks/video-actions/v1\n---\n\n# Human version\n\nThe owner approves the script, then actual storyboard/ingredients and budget. Both authenticated reviews must pass before production.\n\n---\n\n# Agent version\n\nInstalled entry: catalog_fetch type skill make-custom-video here once. Check live custom schemas before writes. Keep this environment; no retired actions, template fallback or reference-as-final import.\n\n## Load the production harness\n\nRequire complete video-production-harness content/scripts/files/version/contentHash. An inline read lists it by slug only: get it with catalog_fetch delivery archive (making needs that shell anyway), or catalog_fetch its slug. Read its orchestrator, phases and editorial/specialist guides. Save safe paths separately with hashes/dependencies. Incomplete/mixed packages block work. Resume recorded packages, inputs and jobs.\n\n## Keep the brief and study the reference\n\nExplicit custom skips video_formats and the picker. Keep requested mechanics/assets; animated text and product photos must not become a creator montage. Implicit references use goose-video to watch and compare template constraints; custom follows no match or their choice.\nRead brand_read (kit, products, rules), then knowledge_search. Use supported claims, real photos/logo. Watch playback, or timed frames spanning the video plus audio/transcript. Record hook, scene order, pace, framing, dialogue, text, product interactions and ending with times/gaps. URLs/thumbnails are not viewing. Missing access needs an accessible source or explicit brief-only choice.\n\n## Create or resume\n\nNew: video_create with brand_id, style custom, request and a required agent-generated UUID idempotency_key; retain products, name, reference_url and campaign. Reuse that key on retries. Resume with video_read the same project: retain plan_revision, custom_review, brand_context, production_manifest and execution; never create a replacement.\nReference resolve accepts direct files; social links cause no automatic paid lookup. If GooseWorks can't download it, upload the file you watched into this project (media_upload scope video_project, media_confirm), wait for media_analyze read to show its duration, then video_change custom reference attach with asset_id and its current_revision as analysis_revision; each refusal names the condition to fix. Skip needs explicit brief-only user_quote. Pending/failed reference blocks approval; a URL is not viewing.\nA resolved reference also needs draft.reference_observation: the project's reference source_url, real duration_s, evidence_mode playback or frames_and_audio, observed_through_s covering 95%, and summary: what happens with times and gaps, as text or a list of {t, what}. Its data binds the script token. Resolve/attach is not watching.\n\n## Save script and ingredients\n\nFollow the harness phases. Discover specialists for the actual scenes; check inputs, outputs, dependencies and managed execution. Reuse suitable product photos and deterministic text animation. Never substitute a template driver. Bind operations to this project and approved scene/script revisions.\nvideo_change needs expected_revision (plan_revision), custom draft and optional script. Bind brand_context_digest to the returned bundle; never author server context. Keep packages outside the review payload. Quote paid operations through the live managed estimate path with real inputs/project. Save preview_estimate/render_estimate for the work still to buy; the server adds committed_credits. Never edit a field the owner approved: the save sets their approval aside until the approved text is back. Missing price/unsupported inputs blocks production; no guessed rates/direct keys.\n\n## Two authenticated owner reviews\n\nShow the complete saved script and exact cumulative budget on the authenticated review card. Only the owner's actual signed button press approves. Chat yes, OAuth, user_quote, widget origin or a copied token is not proof; never manufacture a click or self-approve. Text-only/large reviews use returned authenticated Studio phase buttons and approval_surface URL.\nReread video_read after owner actions. Script approval permits quoted previews only. Save the actual storyboard, playable ingredients and render budget; ingredient approval permits clips. Require current script_approved/ingredients_approved, matching tokens/revision and budget. Changed reviews need renewed approval. A prior yes does not authorize changed work.\n\n## Execute, check and deliver\n\nRequire execution.local from a verified customer shell. No cloud, style-only kit, API host, worker or coworker sandbox. Otherwise report the gap.\nvideo_make needs expected_revision and custom render state. Resume the same render with approved tokens; report running/failed/complete honestly and honor Stop. Save actual prompts, settings, every scene and clip checks through video_change custom production. Watch the full output against the approved script, product and brand; save its full quality report. Completion needs a confirmed upload and rebuild recipe here; pin only the checked complete render. Read back manifest/final before delivery. Repairs need affected assets only and renewed approvals/budget.\n\nRebuild recipe: every input the build reads, uploaded here or a linked Library original, plus recipe/source.json; prove a clean-folder rebuild, then save it as custom production source. The next step gives the format.\n\nUse plain words and returned credits. No invented claims, approvals, assets or passing checks.\n";
+    readonly 'goose-video': "---\nname: goose-video\ndescription: \"Make a video ad in this chat. Study a supplied video reference before comparing styles; follow the next_step in every answer.\"\nversion: 4.4.4\n---\n\n# GooseWorks video\n\nFrom an installed entry, first catalog_fetch type skill goose-video on this same connection and follow that body. An entry already fetched in this run needs no repeat fetch.\n\nFor a video already started, call video_read and follow its next_step.\n\nExplicit custom: skip steps 5–6. Keep their mechanics and assets; research the brand/product, study the reference and follow the custom rule below.\n\nFor other new videos start with get_started (wants video, request). A card is not a style choice or proof of viewing. Follow the reference rule before style selection or a plan.\n\n1. The request. If they haven't said what the video is about, ask in one short question.\n2. The brand. Ask which brand only when brand_read lists several; with one, use it.\n3. The product. Named or none wanted: find its id with brand_read (sections: products), or use []. Otherwise the product card asks: say one line and wait. Without cards, ask which product by name; use the only one, or go on with none. Never ask again after a pick.\n4. Research. Read the brand kit, learnings and the chosen products with brand_read (sections kit and learnings, plus the products' ids), then knowledge_search for them. Note what to say and avoid.\n5. The styles. Call video_formats with the request and the products' ids ([] for none). The customer picks; never recommend one. A style they named that the answer lists as available is their pick. Otherwise, with the style card on screen, say one short line and wait; without one, print the style table from the answer as it is and ask nothing else.\n6. The plan. Call video_create with the picked style and the products; its next_step says what follows.\n\n## A supplied video reference comes first\n\nWatch the actual video before choosing a style or writing a custom plan. Play through it, or inspect timestamped frames spanning it plus its audio/transcript. A URL, title, caption or thumbnail is not a viewing. Record the hook, scene order, pacing, narration/dialogue, framing, product interactions, text and ending, with timestamps and gaps. If access or sampling cannot answer an essential detail, ask for an accessible video or their choice to continue from the brief; never claim it was watched.\nWhen they have not chosen custom, call video_formats after brand research. For plausible matches, catalog_fetch type template returns the style's plan_note, scenes, duration, cast, sound and assets. Compare these against the observed reference and their essential choices. A similar title or high brand ranking does not prove a match. Explain fit or mismatch before they choose. Keep an explicit style choice; do not quietly change constraints or over-adapt it. Reassess a later reference before creating or rewriting a plan.\nIf no available style fits, or they explicitly want custom, catalog_fetch type skill make-custom-video on this connection before a project write and follow it: its full harness, current schemas, two authenticated approvals, cumulative budgets and verified execution. If anything is unavailable, say so and keep the brief; never run a template's renderer and call it custom, import the reference as a finished video, use retired actions or switch environments. Fetching a skill proves no execution capability.\n\n## Rules\n\n- Talk to marketers: what they get and decide, never how it's made.\n- Never mention tools, ids, JSON, commands, software, vendors, models, timings, retries or your checks.\n- Everything happens here: never ask the customer to install anything or use a CLI, slash command or other app, except a video's make command when a next_step gives it (its last line says who starts it; original videos follow their own). Never give any other command, sign-in step or setting, and never ask them a technical question.\n- Money is a credit number exactly as an answer gives it: never dollars or your own math.\n- Follow the next_step in every answer.\n- Card on screen (display_hint \"widget\"): one short line at most, then the wait in its next_step or end your turn; never re-list it or open a question box.\n- Paid work needs the customer's yes to the credit total first, on the card or in chat at the price shown; never decide it for them, and pass on a yes only once they gave it to that total.\n- A yes is to the credit total the customer was shown: their press on the plan card, or their yes in chat sent with that total.\n- Where the app shows no card, print the plan and its price and ask for the customer's yes here; never send them to a website to approve.\n- Never invent an offer, a claim or proof.\n- GooseWorks holds every provider key and bills the work: never ask for a key, token, password or ad account id, and never call a vendor yourself.\n";
+    readonly 'make-custom-video': "---\nname: make-custom-video\ndescription: \"Make a custom video from its brief and studied reference, with authenticated script and ingredient reviews.\"\nversion: 3.5.1\nrequires_skills: [video-production-harness]\nharness_binding: gooseworks/video-actions/v1\n---\n\n# Human version\n\nThe owner approves the script, then actual storyboard/ingredients and budget. Both authenticated reviews must pass before production.\n\n---\n\n# Agent version\n\nInstalled entry: catalog_fetch type skill make-custom-video here once. Check live custom schemas before writes. Keep this environment; no retired actions or reference-as-final import.\n\n## Load the production harness\n\nRequire the complete video-production-harness (content, scripts, files, version, contentHash). An inline read lists it by slug only, so get it with catalog_fetch delivery archive (making needs that shell anyway), unpack it, then read its SKILL.md and follow the reading order it gives (each step file when you reach that step). Save safe paths separately with hashes/dependencies. Incomplete/mixed packages block work. Resume recorded packages, inputs and jobs.\n\n## Keep the brief and study the reference\n\nExplicit custom skips video_formats and the picker. Keep requested mechanics/assets; animated text and product photos must not become a creator montage. Implicit references use goose-video to watch and compare template constraints; custom follows no match or their choice.\nRead brand_read (kit, products, rules), then knowledge_search. Use supported claims, real photos/logo. Watch the reference by playback, or timed frames spanning it plus audio/transcript, and log its beats as the harness says; URLs/thumbnails are not viewing. Missing access needs an accessible source or explicit brief-only choice.\nBefore video_create, catalog_fetch type template for the style nearest the brief and use its plan as your skeleton (harness design brief); with no customer reference, create with its demo video (media_url) as reference_url. Never run a template's renderer or unchanged style and call it custom.\n\n## Create or resume\n\nNew: video_create with brand_id, style custom, request and a required agent-generated UUID idempotency_key; retain products, name, reference_url and campaign. Reuse that key on retries. Resume with video_read the same project: retain plan_revision, custom_review, brand_context, production_manifest and execution; never create a replacement.\nReference resolve takes direct files and starts no paid social lookup. If it fails, attach the file you watched as next_step says; skip needs the customer's brief-only user_quote. A pending/failed reference blocks approval. A resolved one also needs draft.reference_observation (what you saw and heard); it binds the script token.\n\n## Save script and ingredients\n\nFollow the harness phases for the actual scenes; reuse suitable product photos and deterministic text animation. Bind operations to this project and approved scene/script revisions.\nvideo_change needs expected_revision (plan_revision), custom draft and optional script. Bind brand_context_digest to the returned bundle; never author server context. Keep packages outside the review payload. Price each paid request free with estimate_only=1 and save preview_estimate/render_estimate, with its redo allowance, for the work still to buy (harness design brief); the server adds committed_credits. Never edit a field the owner approved: the save sets their approval aside until the approved text is back. Missing price/unsupported inputs blocks production; no guessed rates/direct keys.\n\n## Two authenticated owner reviews\n\nShow the complete saved script and exact cumulative budget on the authenticated review card. Only the owner's actual signed button press approves. Chat yes, OAuth, user_quote, widget origin or a copied token is not proof; never manufacture a click or self-approve. Text-only/large reviews use returned authenticated Studio phase buttons and approval_surface URL.\nReread video_read after owner actions. Script approval permits quoted previews only. Save the actual storyboard, playable ingredients and render budget; ingredient approval permits clips. Require current script_approved/ingredients_approved, matching tokens/revision and budget. Changed reviews need renewed approval. A prior yes does not authorize changed work.\n\n## Execute, check and deliver\n\nWhen the nearest template's parts fit every scene, save the harness's kit block in the ingredients draft; after ingredient approval, follow video_make's next_step (the kit makes, checks and finishes the video; send no render state), then read the final back with video_read. Otherwise require execution.local from a verified customer shell (no cloud, API host, worker or coworker sandbox), or report the gap.\nWhen you build it yourself, video_make needs expected_revision and custom render state. Resume the same render with approved tokens; report running/failed/complete honestly and honor Stop. Report running with a progress_note after each paid piece and at least every 15 minutes. Send each generated clip to custom clip_check and redo a failed one once inside the allowance (harness clip step). Save actual prompts, settings, every scene and clip checks through video_change custom production. Watch the full output against the approved script, product and brand; save its full quality report. Export an MP4 (H.264 + AAC) and complete with output_url set to the output_url media_confirm returned for it; GooseWorks checks the file, so report complete again while it says still checking. Pin only the checked complete render. Read back manifest/final before delivery. Repairs need affected assets only and renewed approvals/budget.\n\nWhen you build it yourself, completion also needs the rebuild recipe saved as custom production source: the harness's references/recipe-format.md and the next_step give the steps.\n\nUse plain words and returned credits. No invented claims, approvals, assets or passing checks.\n";
 };
 export type ServerVideoEntrySlug = keyof typeof SERVER_VIDEO_ENTRY_SKILLS;
 /** renderEntrySkillRules(product) for the products the CLI's own entry skills cover. */

package/dist/skills/server-entry-skills.js

@@ -2,7 +2,7 @@
 // The entry skill text our server builds, copied here because the CLI
 // cannot import gooseworks-app. Do not edit by hand.
 //
-// Source: gooseworks-app akhil/staging-soft-blockers @ bd2f9f70e
+// Source: gooseworks-app akhil/vq-wp7a2-custom-kit-handoff @ 036fa8393
 //   apps/api/src/services/playbooks/entry-skill.ts      buildVideoEntrySkill(slug)
 //   apps/api/src/app-mcp-server/lib/video-policy.ts     renderEntrySkillRules(product)
 //
@@ -29,7 +29,7 @@ From an installed entry, first catalog_fetch type skill goose-video on this same
 
 For a video already started, call video_read and follow its next_step.
 
-Explicit custom: skip steps 5–6. Keep their mechanics and assets; research the brand/product and study the reference for its custom plan. Check current custom support before any write.
+Explicit custom: skip steps 5–6. Keep their mechanics and assets; research the brand/product, study the reference and follow the custom rule below.
 
 For other new videos start with get_started (wants video, request). A card is not a style choice or proof of viewing. Follow the reference rule before style selection or a plan.
 
@@ -44,7 +44,7 @@ For other new videos start with get_started (wants video, request). A card is no
 
 Watch the actual video before choosing a style or writing a custom plan. Play through it, or inspect timestamped frames spanning it plus its audio/transcript. A URL, title, caption or thumbnail is not a viewing. Record the hook, scene order, pacing, narration/dialogue, framing, product interactions, text and ending, with timestamps and gaps. If access or sampling cannot answer an essential detail, ask for an accessible video or their choice to continue from the brief; never claim it was watched.
 When they have not chosen custom, call video_formats after brand research. For plausible matches, catalog_fetch type template returns the style's plan_note, scenes, duration, cast, sound and assets. Compare these against the observed reference and their essential choices. A similar title or high brand ranking does not prove a match. Explain fit or mismatch before they choose. Keep an explicit style choice; do not quietly change constraints or over-adapt it. Reassess a later reference before creating or rewriting a plan.
-If no available style fits, or they explicitly want custom, catalog_fetch type skill make-custom-video on this connection before a project write. Require its complete production-harness dependency and current custom schemas. Follow its separate authenticated script and ingredient approvals, cumulative budgets and verified execution capability. If anything is unavailable, say so and keep the brief; do not substitute a template, import the reference as a finished video, use retired actions or switch environments. Fetching a skill alone proves no execution capability.
+If no available style fits, or they explicitly want custom, catalog_fetch type skill make-custom-video on this connection before a project write and follow it: its full harness, current schemas, two authenticated approvals, cumulative budgets and verified execution. If anything is unavailable, say so and keep the brief; never run a template's renderer and call it custom, import the reference as a finished video, use retired actions or switch environments. Fetching a skill proves no execution capability.
 
 ## Rules
 
@@ -76,27 +76,27 @@ The owner approves the script, then actual storyboard/ingredients and budget. Bo
 
 # Agent version
 
-Installed entry: catalog_fetch type skill make-custom-video here once. Check live custom schemas before writes. Keep this environment; no retired actions, template fallback or reference-as-final import.
+Installed entry: catalog_fetch type skill make-custom-video here once. Check live custom schemas before writes. Keep this environment; no retired actions or reference-as-final import.
 
 ## Load the production harness
 
-Require complete video-production-harness content/scripts/files/version/contentHash. An inline read lists it by slug only: get it with catalog_fetch delivery archive (making needs that shell anyway), or catalog_fetch its slug. Read its orchestrator, phases and editorial/specialist guides. Save safe paths separately with hashes/dependencies. Incomplete/mixed packages block work. Resume recorded packages, inputs and jobs.
+Require the complete video-production-harness (content, scripts, files, version, contentHash). An inline read lists it by slug only, so get it with catalog_fetch delivery archive (making needs that shell anyway), unpack it, then read its SKILL.md and follow the reading order it gives (each step file when you reach that step). Save safe paths separately with hashes/dependencies. Incomplete/mixed packages block work. Resume recorded packages, inputs and jobs.
 
 ## Keep the brief and study the reference
 
 Explicit custom skips video_formats and the picker. Keep requested mechanics/assets; animated text and product photos must not become a creator montage. Implicit references use goose-video to watch and compare template constraints; custom follows no match or their choice.
-Read brand_read (kit, products, rules), then knowledge_search. Use supported claims, real photos/logo. Watch playback, or timed frames spanning the video plus audio/transcript. Record hook, scene order, pace, framing, dialogue, text, product interactions and ending with times/gaps. URLs/thumbnails are not viewing. Missing access needs an accessible source or explicit brief-only choice.
+Read brand_read (kit, products, rules), then knowledge_search. Use supported claims, real photos/logo. Watch the reference by playback, or timed frames spanning it plus audio/transcript, and log its beats as the harness says; URLs/thumbnails are not viewing. Missing access needs an accessible source or explicit brief-only choice.
+Before video_create, catalog_fetch type template for the style nearest the brief and use its plan as your skeleton (harness design brief); with no customer reference, create with its demo video (media_url) as reference_url. Never run a template's renderer or unchanged style and call it custom.
 
 ## Create or resume
 
 New: video_create with brand_id, style custom, request and a required agent-generated UUID idempotency_key; retain products, name, reference_url and campaign. Reuse that key on retries. Resume with video_read the same project: retain plan_revision, custom_review, brand_context, production_manifest and execution; never create a replacement.
-Reference resolve accepts direct files; social links cause no automatic paid lookup. If GooseWorks can't download it, upload the file you watched into this project (media_upload scope video_project, media_confirm), wait for media_analyze read to show its duration, then video_change custom reference attach with asset_id and its current_revision as analysis_revision; each refusal names the condition to fix. Skip needs explicit brief-only user_quote. Pending/failed reference blocks approval; a URL is not viewing.
-A resolved reference also needs draft.reference_observation: the project's reference source_url, real duration_s, evidence_mode playback or frames_and_audio, observed_through_s covering 95%, and summary: what happens with times and gaps, as text or a list of {t, what}. Its data binds the script token. Resolve/attach is not watching.
+Reference resolve takes direct files and starts no paid social lookup. If it fails, attach the file you watched as next_step says; skip needs the customer's brief-only user_quote. A pending/failed reference blocks approval. A resolved one also needs draft.reference_observation (what you saw and heard); it binds the script token.
 
 ## Save script and ingredients
 
-Follow the harness phases. Discover specialists for the actual scenes; check inputs, outputs, dependencies and managed execution. Reuse suitable product photos and deterministic text animation. Never substitute a template driver. Bind operations to this project and approved scene/script revisions.
-video_change needs expected_revision (plan_revision), custom draft and optional script. Bind brand_context_digest to the returned bundle; never author server context. Keep packages outside the review payload. Quote paid operations through the live managed estimate path with real inputs/project. Save preview_estimate/render_estimate for the work still to buy; the server adds committed_credits. Never edit a field the owner approved: the save sets their approval aside until the approved text is back. Missing price/unsupported inputs blocks production; no guessed rates/direct keys.
+Follow the harness phases for the actual scenes; reuse suitable product photos and deterministic text animation. Bind operations to this project and approved scene/script revisions.
+video_change needs expected_revision (plan_revision), custom draft and optional script. Bind brand_context_digest to the returned bundle; never author server context. Keep packages outside the review payload. Price each paid request free with estimate_only=1 and save preview_estimate/render_estimate, with its redo allowance, for the work still to buy (harness design brief); the server adds committed_credits. Never edit a field the owner approved: the save sets their approval aside until the approved text is back. Missing price/unsupported inputs blocks production; no guessed rates/direct keys.
 
 ## Two authenticated owner reviews
 
@@ -105,10 +105,10 @@ Reread video_read after owner actions. Script approval permits quoted previews o
 
 ## Execute, check and deliver
 
-Require execution.local from a verified customer shell. No cloud, style-only kit, API host, worker or coworker sandbox. Otherwise report the gap.
-video_make needs expected_revision and custom render state. Resume the same render with approved tokens; report running/failed/complete honestly and honor Stop. Save actual prompts, settings, every scene and clip checks through video_change custom production. Watch the full output against the approved script, product and brand; save its full quality report. Completion needs a confirmed upload and rebuild recipe here; pin only the checked complete render. Read back manifest/final before delivery. Repairs need affected assets only and renewed approvals/budget.
+When the nearest template's parts fit every scene, save the harness's kit block in the ingredients draft; after ingredient approval, follow video_make's next_step (the kit makes, checks and finishes the video; send no render state), then read the final back with video_read. Otherwise require execution.local from a verified customer shell (no cloud, API host, worker or coworker sandbox), or report the gap.
+When you build it yourself, video_make needs expected_revision and custom render state. Resume the same render with approved tokens; report running/failed/complete honestly and honor Stop. Report running with a progress_note after each paid piece and at least every 15 minutes. Send each generated clip to custom clip_check and redo a failed one once inside the allowance (harness clip step). Save actual prompts, settings, every scene and clip checks through video_change custom production. Watch the full output against the approved script, product and brand; save its full quality report. Export an MP4 (H.264 + AAC) and complete with output_url set to the output_url media_confirm returned for it; GooseWorks checks the file, so report complete again while it says still checking. Pin only the checked complete render. Read back manifest/final before delivery. Repairs need affected assets only and renewed approvals/budget.
 
-Rebuild recipe: every input the build reads, uploaded here or a linked Library original, plus recipe/source.json; prove a clean-folder rebuild, then save it as custom production source. The next step gives the format.
+When you build it yourself, completion also needs the rebuild recipe saved as custom production source: the harness's references/recipe-format.md and the next_step give the steps.
 
 Use plain words and returned credits. No invented claims, approvals, assets or passing checks.
 `,

package/dist/skills/server-entry-skills.js.map

@@ -1 +1 @@
-{"version":3,"file":"server-entry-skills.js","sourceRoot":"","sources":["../../src/skills/server-entry-skills.ts"],"names":[],"mappings":";AAAA,sEAAsE;AACtE,qDAAqD;AACrD,EAAE;AACF,iEAAiE;AACjE,mFAAmF;AACnF,uFAAuF;AACvF,EAAE;AACF,+EAA+E;AAC/E,wEAAwE;AACxE,6EAA6E;AAC7E,6EAA6E;AAC7E,sEAAsE;AACtE,iEAAiE;;;AAEjE,qDAAqD;AACxC,QAAA,uBAAuB,GAAG,OAAO,CAAC;AAElC,QAAA,yBAAyB,GAAG;IACvC,aAAa,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA0ChB;IACC,mBAAmB,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAmDtB;CACS,CAAC;AAIX,wFAAwF;AAC3E,QAAA,wBAAwB,GAAG;IACtC,GAAG,EAAE;;;;;;;;mJAQ4I;IACjJ,GAAG,EAAE;;;;;;;;;mJAS4I;IACjJ,MAAM,EAAE;;;;;;;;;mJASyI;CACzI,CAAC"}
+{"version":3,"file":"server-entry-skills.js","sourceRoot":"","sources":["../../src/skills/server-entry-skills.ts"],"names":[],"mappings":";AAAA,sEAAsE;AACtE,qDAAqD;AACrD,EAAE;AACF,uEAAuE;AACvE,mFAAmF;AACnF,uFAAuF;AACvF,EAAE;AACF,+EAA+E;AAC/E,wEAAwE;AACxE,6EAA6E;AAC7E,6EAA6E;AAC7E,sEAAsE;AACtE,iEAAiE;;;AAEjE,qDAAqD;AACxC,QAAA,uBAAuB,GAAG,OAAO,CAAC;AAElC,QAAA,yBAAyB,GAAG;IACvC,aAAa,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA0ChB;IACC,mBAAmB,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAmDtB;CACS,CAAC;AAIX,wFAAwF;AAC3E,QAAA,wBAAwB,GAAG;IACtC,GAAG,EAAE;;;;;;;;mJAQ4I;IACjJ,GAAG,EAAE;;;;;;;;;mJAS4I;IACjJ,MAAM,EAAE;;;;;;;;;mJASyI;CACzI,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gooseworks",
-  "version": "0.5.9-dev.58.1cb81d21",
+  "version": "0.5.9-dev.59.2b371ca1",
   "description": "GooseWorks CLI — give your coding agent real data tools",
   "bin": {
     "gooseworks": "./dist/index.js"
@@ -72,7 +72,7 @@
   "homepage": "https://gooseworks.ai",
   "author": "GooseWorks <help@gooseworks.ai>",
   "gooseworksRelease": {
-    "commit": "1cb81d21322a1ab4c4b53c001a4053c736a46c83",
+    "commit": "2b371ca1b17eae291d8782c35e262c1a76f92e2b",
     "branch": "dev"
   }
 }

package/skills/goose-video/SKILL.md

@@ -10,7 +10,7 @@ From an installed entry, first catalog_fetch type skill goose-video on this same
 
 For a video already started, call video_read and follow its next_step.
 
-Explicit custom: skip steps 5–6. Keep their mechanics and assets; research the brand/product and study the reference for its custom plan. Check current custom support before any write.
+Explicit custom: skip steps 5–6. Keep their mechanics and assets; research the brand/product, study the reference and follow the custom rule below.
 
 For other new videos start with get_started (wants video, request). A card is not a style choice or proof of viewing. Follow the reference rule before style selection or a plan.
 
@@ -25,7 +25,7 @@ For other new videos start with get_started (wants video, request). A card is no
 
 Watch the actual video before choosing a style or writing a custom plan. Play through it, or inspect timestamped frames spanning it plus its audio/transcript. A URL, title, caption or thumbnail is not a viewing. Record the hook, scene order, pacing, narration/dialogue, framing, product interactions, text and ending, with timestamps and gaps. If access or sampling cannot answer an essential detail, ask for an accessible video or their choice to continue from the brief; never claim it was watched.
 When they have not chosen custom, call video_formats after brand research. For plausible matches, catalog_fetch type template returns the style's plan_note, scenes, duration, cast, sound and assets. Compare these against the observed reference and their essential choices. A similar title or high brand ranking does not prove a match. Explain fit or mismatch before they choose. Keep an explicit style choice; do not quietly change constraints or over-adapt it. Reassess a later reference before creating or rewriting a plan.
-If no available style fits, or they explicitly want custom, catalog_fetch type skill make-custom-video on this connection before a project write. Require its complete production-harness dependency and current custom schemas. Follow its separate authenticated script and ingredient approvals, cumulative budgets and verified execution capability. If anything is unavailable, say so and keep the brief; do not substitute a template, import the reference as a finished video, use retired actions or switch environments. Fetching a skill alone proves no execution capability.
+If no available style fits, or they explicitly want custom, catalog_fetch type skill make-custom-video on this connection before a project write and follow it: its full harness, current schemas, two authenticated approvals, cumulative budgets and verified execution. If anything is unavailable, say so and keep the brief; never run a template's renderer and call it custom, import the reference as a finished video, use retired actions or switch environments. Fetching a skill proves no execution capability.
 
 ## Rules
 

package/skills/make-custom-video/SKILL.md

@@ -14,27 +14,27 @@ The owner approves the script, then actual storyboard/ingredients and budget. Bo
 
 # Agent version
 
-Installed entry: catalog_fetch type skill make-custom-video here once. Check live custom schemas before writes. Keep this environment; no retired actions, template fallback or reference-as-final import.
+Installed entry: catalog_fetch type skill make-custom-video here once. Check live custom schemas before writes. Keep this environment; no retired actions or reference-as-final import.
 
 ## Load the production harness
 
-Require complete video-production-harness content/scripts/files/version/contentHash. An inline read lists it by slug only: get it with catalog_fetch delivery archive (making needs that shell anyway), or catalog_fetch its slug. Read its orchestrator, phases and editorial/specialist guides. Save safe paths separately with hashes/dependencies. Incomplete/mixed packages block work. Resume recorded packages, inputs and jobs.
+Require the complete video-production-harness (content, scripts, files, version, contentHash). An inline read lists it by slug only, so get it with catalog_fetch delivery archive (making needs that shell anyway), unpack it, then read its SKILL.md and follow the reading order it gives (each step file when you reach that step). Save safe paths separately with hashes/dependencies. Incomplete/mixed packages block work. Resume recorded packages, inputs and jobs.
 
 ## Keep the brief and study the reference
 
 Explicit custom skips video_formats and the picker. Keep requested mechanics/assets; animated text and product photos must not become a creator montage. Implicit references use goose-video to watch and compare template constraints; custom follows no match or their choice.
-Read brand_read (kit, products, rules), then knowledge_search. Use supported claims, real photos/logo. Watch playback, or timed frames spanning the video plus audio/transcript. Record hook, scene order, pace, framing, dialogue, text, product interactions and ending with times/gaps. URLs/thumbnails are not viewing. Missing access needs an accessible source or explicit brief-only choice.
+Read brand_read (kit, products, rules), then knowledge_search. Use supported claims, real photos/logo. Watch the reference by playback, or timed frames spanning it plus audio/transcript, and log its beats as the harness says; URLs/thumbnails are not viewing. Missing access needs an accessible source or explicit brief-only choice.
+Before video_create, catalog_fetch type template for the style nearest the brief and use its plan as your skeleton (harness design brief); with no customer reference, create with its demo video (media_url) as reference_url. Never run a template's renderer or unchanged style and call it custom.
 
 ## Create or resume
 
 New: video_create with brand_id, style custom, request and a required agent-generated UUID idempotency_key; retain products, name, reference_url and campaign. Reuse that key on retries. Resume with video_read the same project: retain plan_revision, custom_review, brand_context, production_manifest and execution; never create a replacement.
-Reference resolve accepts direct files; social links cause no automatic paid lookup. If GooseWorks can't download it, upload the file you watched into this project (media_upload scope video_project, media_confirm), wait for media_analyze read to show its duration, then video_change custom reference attach with asset_id and its current_revision as analysis_revision; each refusal names the condition to fix. Skip needs explicit brief-only user_quote. Pending/failed reference blocks approval; a URL is not viewing.
-A resolved reference also needs draft.reference_observation: the project's reference source_url, real duration_s, evidence_mode playback or frames_and_audio, observed_through_s covering 95%, and summary: what happens with times and gaps, as text or a list of {t, what}. Its data binds the script token. Resolve/attach is not watching.
+Reference resolve takes direct files and starts no paid social lookup. If it fails, attach the file you watched as next_step says; skip needs the customer's brief-only user_quote. A pending/failed reference blocks approval. A resolved one also needs draft.reference_observation (what you saw and heard); it binds the script token.
 
 ## Save script and ingredients
 
-Follow the harness phases. Discover specialists for the actual scenes; check inputs, outputs, dependencies and managed execution. Reuse suitable product photos and deterministic text animation. Never substitute a template driver. Bind operations to this project and approved scene/script revisions.
-video_change needs expected_revision (plan_revision), custom draft and optional script. Bind brand_context_digest to the returned bundle; never author server context. Keep packages outside the review payload. Quote paid operations through the live managed estimate path with real inputs/project. Save preview_estimate/render_estimate for the work still to buy; the server adds committed_credits. Never edit a field the owner approved: the save sets their approval aside until the approved text is back. Missing price/unsupported inputs blocks production; no guessed rates/direct keys.
+Follow the harness phases for the actual scenes; reuse suitable product photos and deterministic text animation. Bind operations to this project and approved scene/script revisions.
+video_change needs expected_revision (plan_revision), custom draft and optional script. Bind brand_context_digest to the returned bundle; never author server context. Keep packages outside the review payload. Price each paid request free with estimate_only=1 and save preview_estimate/render_estimate, with its redo allowance, for the work still to buy (harness design brief); the server adds committed_credits. Never edit a field the owner approved: the save sets their approval aside until the approved text is back. Missing price/unsupported inputs blocks production; no guessed rates/direct keys.
 
 ## Two authenticated owner reviews
 
@@ -43,9 +43,9 @@ Reread video_read after owner actions. Script approval permits quoted previews o
 
 ## Execute, check and deliver
 
-Require execution.local from a verified customer shell. No cloud, style-only kit, API host, worker or coworker sandbox. Otherwise report the gap.
-video_make needs expected_revision and custom render state. Resume the same render with approved tokens; report running/failed/complete honestly and honor Stop. Save actual prompts, settings, every scene and clip checks through video_change custom production. Watch the full output against the approved script, product and brand; save its full quality report. Completion needs a confirmed upload and rebuild recipe here; pin only the checked complete render. Read back manifest/final before delivery. Repairs need affected assets only and renewed approvals/budget.
+When the nearest template's parts fit every scene, save the harness's kit block in the ingredients draft; after ingredient approval, follow video_make's next_step (the kit makes, checks and finishes the video; send no render state), then read the final back with video_read. Otherwise require execution.local from a verified customer shell (no cloud, API host, worker or coworker sandbox), or report the gap.
+When you build it yourself, video_make needs expected_revision and custom render state. Resume the same render with approved tokens; report running/failed/complete honestly and honor Stop. Report running with a progress_note after each paid piece and at least every 15 minutes. Send each generated clip to custom clip_check and redo a failed one once inside the allowance (harness clip step). Save actual prompts, settings, every scene and clip checks through video_change custom production. Watch the full output against the approved script, product and brand; save its full quality report. Export an MP4 (H.264 + AAC) and complete with output_url set to the output_url media_confirm returned for it; GooseWorks checks the file, so report complete again while it says still checking. Pin only the checked complete render. Read back manifest/final before delivery. Repairs need affected assets only and renewed approvals/budget.
 
-Rebuild recipe: every input the build reads, uploaded here or a linked Library original, plus recipe/source.json; prove a clean-folder rebuild, then save it as custom production source. The next step gives the format.
+When you build it yourself, completion also needs the rebuild recipe saved as custom production source: the harness's references/recipe-format.md and the next_step give the steps.
 
 Use plain words and returned credits. No invented claims, approvals, assets or passing checks.

package/skills/manifest.json

@@ -1,12 +1,12 @@
 {
   "package": "gooseworks",
-  "version": "0.5.9-dev.58.1cb81d21",
+  "version": "0.5.9-dev.59.2b371ca1",
   "entries": {
     "gooseworks": "2b5507cd937b2e5371bc1645530a13bda219914c9ecc4b14ea3871a9f43b95db",
     "goose-ads": "ee52c2e83edc196bc22121e089b9bce526b789ce762935c35e1c989f918c6569",
-    "goose-video": "3de2f0aa3d38afb462d59dbe7778e2395adc551283b6ec19a90c63e6f6b6008d",
+    "goose-video": "b4cb140d589292b76ecec06eb04868c9f16882c5198ffe4729d92e2cfe652186",
     "goose-video-local": "3c8c8aecbfe3d8e58e6393e28dc67dc36f9d4b8dcfb1603f8bc442786ce839ee",
-    "make-custom-video": "fe7bd6d35023d372370a1ff66577ba4ae0bbc76451553dac20ed16dd895e51cf",
+    "make-custom-video": "47b11a5664f7a9d02ee83d6c7cfb2f3526a6045fd6bedddae6e888a32132882b",
     "goose-product-photos": "a8d04bcb320fb71885ff8df90cb32093b94c38d7cbf70f83d9275c6ff7e0be29"
   }
 }