← falCONTENT HISTORY

Update to fal

Snapshot Nov 14, 2026 · 00:10 UTC · version 1.3.2

Collection source: downloaded plugin package. These snapshots do not have a confirmed matching collection source. Differences in file lists alone do not establish changes to the package.

WHAT CHANGED

Instructions updated for fan-cam

Skill instructions — Now says: Bundled guidance is fallback craft. For inputs, recommend_model's `input_contract` and live validation at submission are authoritative. See [the shared...

Observed in instructions or declared skills. Runtime behavior has not been tested.

Save fal and check future changes in Following.

Up to 3 plugins free · no card required.

All fields and original values.

Skill instructions

Before

Sep 30, 2026 · 22:46 UTC

3. Use image-to-video from that approved frame. ## Model Choices - Identity-preserving frame: `openai/gpt-image-2/edit` or another strong edit endpoint after schema inspection. - Broadcast video: Kling v3 image-to-video or Seedance ima...

After

Nov 14, 2026 · 00:10 UTC

Bundled guidance is fallback craft. For inputs, recommend_model's `input_contract` and live validation at submission are authoritative. See [the shared production contract](../shared/production-contract.md). 3. Use image-to-video from th...

Compare saved observations

Download comparison JSON

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -9,6 +9,10 @@
 
 # Fan Cam
 
+Bundled guidance is fallback craft. For inputs, recommend_model's
+`input_contract` and live validation at submission are authoritative. See
+[the shared production contract](../shared/production-contract.md).
+
 Use when the user wants a personalized spectator video that feels like a live
 sports broadcast cutaway. Execution runs through fal.ai MCP tools.
 
@@ -24,7 +28,7 @@
 
 1. Upload the portrait reference if local.
 2. Generate a broadcast-still frame that preserves identity.
-3. Use image-to-video from that approved frame.
+3. Use image-to-video from that frame in the same turn.
 4. Add sports broadcast details in prompt only when they are generic or
    user-provided. Do not invent real network marks or false event claims.
 5. Review identity, crowd plausibility, scoreboard/text artifacts, and motion.
@@ -37,12 +41,13 @@
 - Avoid fake readable scoreboards unless the user supplies exact text.
 - Keep motion natural: cheering, gasping, standing, waving, phone held up.
 
-## Model Choices
+## Model selection
 
-- Identity-preserving frame: `openai/gpt-image-2/edit` or another strong edit
-  endpoint after schema inspection.
-- Broadcast video: Kling v3 image-to-video or Seedance image-to-video,
-  depending on available controls and quality target.
+Use `model-routing` with this brief and actual references, then submit the
+selected `next_call`. Keep a model the user named. Bundled family suggestions
+must not override the current task-specific recommendation. Inspect one result
+before dependent continuity shots; submit independent variations together with
+`submit_jobs`. Preserve originals through any permitted repair.
 
 ## Quality Bar
 
Full technical diff · 1 changed fields

changed /skill_md_contents

BEFORE
"---\nname: fan-cam\ndescription: >\n  Create personalized live sports broadcast fan-cam videos with fal.ai. Use\n  for realistic spectator cutaways, stadium crowd reactions, broadcast-style\n  screenshots, scoreboard overlays, TV bugs, and identity-preserving fan\n  reaction videos from a user photo.\n---\n\n# Fan Cam\n\nUse when the user wants a personalized spectator video that feels like a live\nsports broadcast cutaway. Execution runs through fal.ai MCP tools.\n\n## Required Inputs\n\n- User photo or approved portrait reference.\n- Sport, team/event context, venue, and crowd mood.\n- Reaction: shocked, celebrating, nervous, chanting, laughing, emotional.\n- Broadcast format: close-up cutaway, crowd pan, scoreboard moment, replay.\n- Duration and crop, usually 9:16 or 16:9.\n\n## Production Pattern\n\n1. Upload the portrait reference if local.\n2. Generate a broadcast-still frame that preserves identity.\n3. Use image-to-video from that approved frame.\n4. Add sports broadcast details in prompt only when they are generic or\n   user-provided. Do not invent real network marks or false event claims.\n5. Review identity, crowd plausibility, scoreboard/text artifacts, and motion.\n\n## Prompt Guardrails\n\n- Keep the person recognizable from the reference.\n- Use generic broadcast styling unless the user supplies exact rights-safe\n  graphics.\n- Avoid fake readable scoreboards unless the user supplies exact text.\n- Keep motion natural: cheering, gasping, standing, waving, phone held up.\n\n## Model Choices\n\n- Identity-preserving frame: `openai/gpt-image-2/edit` or another strong edit\n  endpoint after schema inspection.\n- Broadcast video: Kling v3 image-to-video or Seedance image-to-video,\n  depending on available controls and quality target.\n\n## Quality Bar\n\n- Person remains recognizable.\n- Broadcast frame feels live, not posed studio content.\n- Crowd and venue support the sport.\n- Any text/logos are absent, generic, or explicitly provided.\n"
AFTER
"---\nname: fan-cam\ndescription: >\n  Create personalized live sports broadcast fan-cam videos with fal.ai. Use\n  for realistic spectator cutaways, stadium crowd reactions, broadcast-style\n  screenshots, scoreboard overlays, TV bugs, and identity-preserving fan\n  reaction videos from a user photo.\n---\n\n# Fan Cam\n\nBundled guidance is fallback craft. For inputs, recommend_model's\n`input_contract` and live validation at submission are authoritative. See\n[the shared production contract](../shared/production-contract.md).\n\nUse when the user wants a personalized spectator video that feels like a live\nsports broadcast cutaway. Execution runs through fal.ai MCP tools.\n\n## Required Inputs\n\n- User photo or approved portrait reference.\n- Sport, team/event context, venue, and crowd mood.\n- Reaction: shocked, celebrating, nervous, chanting, laughing, emotional.\n- Broadcast format: close-up cutaway, crowd pan, scoreboard moment, replay.\n- Duration and crop, usually 9:16 or 16:9.\n\n## Production Pattern\n\n1. Upload the portrait reference if local.\n2. Generate a broadcast-still frame that preserves identity.\n3. Use image-to-video from that frame in the same turn.\n4. Add sports broadcast details in prompt only when they are generic or\n   user-provided. Do not invent real network marks or false event claims.\n5. Review identity, crowd plausibility, scoreboard/text artifacts, and motion.\n\n## Prompt Guardrails\n\n- Keep the person recognizable from the reference.\n- Use generic broadcast styling unless the user supplies exact rights-safe\n  graphics.\n- Avoid fake readable scoreboards unless the user supplies exact text.\n- Keep motion natural: cheering, gasping, standing, waving, phone held up.\n\n## Model selection\n\nUse `model-routing` with this brief and actual references, then submit the\nselected `next_call`. Keep a model the user named. Bundled family suggestions\nmust not override the current task-specific recommendation. Inspect one result\nbefore dependent continuity shots; submit independent variations together with\n`submit_jobs`. Preserve originals through any permitted repair.\n\n## Quality Bar\n\n- Person remains recognizable.\n- Broadcast frame feels live, not posed studio content.\n- Crowd and venue support the sport.\n- Any text/logos are absent, generic, or explicitly provided.\n"
Full snapshot data
{
  "description": "Create personalized live sports broadcast fan-cam videos with fal.ai. Use for realistic spectator cutaways, stadium crowd reactions, broadcast-style screenshots, scoreboard overlays, TV bugs, and identity-preserving fan reaction videos from a user photo.\n",
  "included_files": [],
  "name": "fan-cam",
  "skill_md_contents": "---\nname: fan-cam\ndescription: >\n  Create personalized live sports broadcast fan-cam videos with fal.ai. Use\n  for realistic spectator cutaways, stadium crowd reactions, broadcast-style\n  screenshots, scoreboard overlays, TV bugs, and identity-preserving fan\n  reaction videos from a user photo.\n---\n\n# Fan Cam\n\nBundled guidance is fallback craft. For inputs, recommend_model's\n`input_contract` and live validation at submission are authoritative. See\n[the shared production contract](../shared/production-contract.md).\n\nUse when the user wants a personalized spectator video that feels like a live\nsports broadcast cutaway. Execution runs through fal.ai MCP tools.\n\n## Required Inputs\n\n- User photo or approved portrait reference.\n- Sport, team/event context, venue, and crowd mood.\n- Reaction: shocked, celebrating, nervous, chanting, laughing, emotional.\n- Broadcast format: close-up cutaway, crowd pan, scoreboard moment, replay.\n- Duration and crop, usually 9:16 or 16:9.\n\n## Production Pattern\n\n1. Upload the portrait reference if local.\n2. Generate a broadcast-still frame that preserves identity.\n3. Use image-to-video from that frame in the same turn.\n4. Add sports broadcast details in prompt only when they are generic or\n   user-provided. Do not invent real network marks or false event claims.\n5. Review identity, crowd plausibility, scoreboard/text artifacts, and motion.\n\n## Prompt Guardrails\n\n- Keep the person recognizable from the reference.\n- Use generic broadcast styling unless the user supplies exact rights-safe\n  graphics.\n- Avoid fake readable scoreboards unless the user supplies exact text.\n- Keep motion natural: cheering, gasping, standing, waving, phone held up.\n\n## Model selection\n\nUse `model-routing` with this brief and actual references, then submit the\nselected `next_call`. Keep a model the user named. Bundled family suggestions\nmust not override the current task-specific recommendation. Inspect one result\nbefore dependent continuity shots; submit independent variations together with\n`submit_jobs`. Preserve originals through any permitted repair.\n\n## Quality Bar\n\n- Person remains recognizable.\n- Broadcast frame feels live, not posed studio content.\n- Crowd and venue support the sport.\n- Any text/logos are absent, generic, or explicitly provided.\n"
}

SHA-256 of public snapshot: 0c91354ed079c77bafdcd3279f80eab3cc05d6defb272152444c6a32a168aeb9