Agent skills
A skill is a short, opinionated instruction file an agent reads before it acts. These cover the judgement calls this API asks for — which model fits a brief, what a generation will cost, how to write a prompt that produces the shot someone described, and what to do when a call fails.
They ship with the aicontentdrop package and are served here so an agent can fetch one without installing anything.
| Skill | What it decides |
|---|---|
| Picking an AI video model (and knowing what it costs) | Choose the right AI video model for a brief and quote its credit cost before generating, using the AI Content Drop catalogue. Use when asked to generate a video and you must decide which model, when someone asks what a batch of clips will cost, or when a model id is rejected as unknown. |
| Failing well against the AI Content Drop API | Read AI Content Drop API failures correctly and decide whether to retry, change the request, or stop. Covers the error envelope and its 21 codes, rate limits and Retry-After, idempotency conflicts, and the two response shapes the generation endpoints can return. Use when a call to aicontentdrop.com fails, returns 4xx or 5xx, is rate limited, or when writing a retry loop against the API. |
| Running many generations without melting the rate limit | Run many AI Content Drop reads or generations efficiently instead of looping one HTTP call at a time. Covers POST /v1/batch for reads, POST /v1/models/cost for bulk quoting, idempotency keys, and the submit-and-poll loop for generations. Use when handling a storyboard, a shot list, a spreadsheet of prompts, or any job that means more than two or three calls. |
| Planning and rendering on a board | Plan and render content on an AI Content Drop board (visual canvas) from an agent. Covers picking or creating a board, laying out text, media and generation cards, quoting with estimate_credit_cost before run_generation_card, asking once per task how spending is approved, never tracking the board revision yourself, the get_canvas_link fallback for hosts that cannot render the board inline, and the REVISION_CONFLICT, CARD_BUSY and CANVAS_FULL refusals. Use when asked to plan a campaign visually, put ideas or assets on a board, or render the cards on one. |
| From competitor ads to a brief you can shoot | Research the ads competitors are running and turn a winner into a brief, from an agent, with AI Content Drop's research tools. Covers search_competitor_ads on Meta or TikTok, never passing sample rows off as real ads, saving an ad, decoding it into a reusable formula with a quote first, and placing it on a board so it can be remade. Use when asked what competitors are running, which hooks work in a category, to break down an ad, or to make an ad "like this one". |
| What it costs, and how to say so before spending it | Answer what an AI Content Drop generation costs, what a plan includes, and whether a user can afford a job before running it. Covers flat and per-second prices, the free tier, where plan details live, checking a balance, and the charge-on-success rule. Use when asked about pricing, credits, billing, "can I afford this", "how much will this cost", or when a call returns insufficient_credits. |
| A short drama, from premise to finished episode | Make a short vertical drama with AI Content Drop's Drama Studio from an agent, from a premise to a finished, cut episode. Covers start_drama, get_drama, list_dramas, revise_drama, review_drama_draft, quote_drama_episode, produce_drama_episode and assemble_drama_episode; following a draft that takes minutes; reading the script back before anything renders; showing the episode total before rendering; and handing over the Drama Studio page when the drama tools are not listed. Use when asked for a short drama, a mini-series, a vertical episode, a soap-style story, or a scripted series of clips with recurring characters. |
| Generating images, and not overpaying for drafts | Generate images through the AI Content Drop API and choose between multiple leading image models by cost and job. Covers POST /v1/generate/image, picking a model for drafts versus finals, the sandbox rehearsal, and using a generated image as the first frame of a video. Use when asked for an image, a thumbnail, a product still, a poster, concept art, or a reference frame to animate. |
| Making a finished ad with the studio tools | Make a finished video ad through the AI Content Drop studio tools from an agent. Covers gathering the brief first, bringing assets in by URL, quoting with estimate_ad_cost before generate_ad, asking once before any credits are spent, polling get_ad_job every 15 seconds, the consent rule for UGC avatar videos, and what to say when a plan is needed. Use when asked to make an ad, a product video, a UGC-style clip, or an avatar presenter from a brief. |
| Driving AI Content Drop from a self-built client | Drive AI Content Drop from a host that writes its own client against the MCP server rather than reading a config file. Covers which tool answers which question, quoting with estimate_credit_cost before any generation, submitting and polling instead of waiting inside a tool call, the drop and drama tools, and repeating refusal codes verbatim. Use when connecting AI Content Drop to a consumer agent as a custom integration, when a saved integration needs to be re-taught the call order, or when a generation call times out, is refused, or is about to be retried. |
| Writing a short drama people keep swiping for | Write a short vertical drama that holds a phone audience, as text, with a premise with built-in conflict, an episode outline with a hook in the first seconds, one visible power move per scene, a status reversal, a shot for every beat and an ending that demands the next episode. Produces a logline, cast, and scene-by-scene beats with shots and dialogue sized to 15, 30, 60 or 90 second episodes, ready to paste into AI Content Drop's Drama Studio. Use when asked to write, outline, fix or punch up a short drama, a mini-series, a vertical episode, a soap-style story, or a premise for one. Text advice only; it renders nothing. |
| Handing the whole job to Drop, the video agent | Hand a whole piece of video work to Drop, the AI Content Drop video agent, from another agent. One brief is carried through research, concepts, a storyboard, renders and assembly, and every step that charges credits stops for an approval. Covers start_video_agent, message_video_agent, get_video_agent_run, approve_video_agent_spend and list_video_agent_runs, relaying the drop's clarifying questions, following a turn that outlives the call, approving the exact quote, and what to do before the person has signed in. Use when asked for a campaign, several videos from one brief, an ad built from competitor research, or "just make the whole thing". |
| Writing a video prompt that survives the model | Write a prompt that actually produces the shot someone described, for any AI Content Drop video model. Covers the shot-first structure, camera and lighting language models respond to, why text-to-video cannot preserve a specific subject, when to use a reference frame instead, and how to iterate without wasting credits. Use when asked to generate a video, turn a brief or script into a prompt, fix a generation that came out wrong, or animate an existing image. |
Using them
Any agent that can read a URL can use a skill: point it at the markdown and let it follow the instructions.
curl https://aicontentdrop.com/skills/ai-video-model-picker/SKILL.md
Enumerate them from /skills/index.json, which carries a description and URL for each. Installing the plugin brings all of them along — see the install guide.
These are guides, not the product
The skills teach craft. Doing the work needs the API: the tools are described at /docs/mcp and the REST surface at /docs/api. Browsing the catalogue and pricing a job need no credential.