agentleFS
MicrockMicrock/kagi-cli/skills/kagi-ai/SKILL.md

kagi-ai

Skill190 starsChanged 8 months agoMITView on GitHub

Use Kagi's paid AI allowance for reasoning, coding, research, file analysis, structured generation, or model comparison. Use when a task should spend Assistant value instead of only searching the web or transforming one known page.

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 7 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use only model slugs and assistant names returned by the account.

    line 37
  2. Distinctive:

    Use a closed model call for reasoning that needs no current information:

    line 44
  3. Distinctive:

    Use a built-in contract for reusable decisions, plans, and checklists:

    line 93
  4. Very specific:

    Use --contract-file PATH when the consumer needs a different small JSON shape.

    line 102
  5. Distinctive:

    Use subscriber Summarizer or Translate for direct transformations when the active plan includes them.

    line 107
  6. Distinctive:

    Use the account default or a lower-cost model for simple work.

    line 109
  7. Distinctive:

    Keep prompts specific and attach only relevant files.

    line 111
  8. Distinctive:

    Start a new thread for an unrelated task.

    line 112
  9. Distinctive:

    every explicit model or assistant came from live account discovery.

    line 121

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Kagi AI

Spend the Assistant allowance on reasoning and synthesis. When the active plan includes unlimited Summarizer or Translate usage, use those tools for direct transformations and reserve the allowance for harder work.

Choose the lane

NeedLane
Find or verify sourcesLoad kagi-usage and search
Summarize one URL or textkagi summarize --subscriber
Translate supplied textkagi translate
Get a fast cited answerkagi quick
Reason, code, synthesize, or analyze fileskagi assistant
Run multi-step research or tool useAn account-visible Research assistant
Manage threads or custom assistantsLoad kagi-assistant

The Assistant allowance is separate from API credit. A plan's monthly value sets its token allowance across Assistant models. Premium models, web access, large files, and long thread histories consume more of that allowance.

Start with live account state

kagi auth status
kagi assistant models
kagi assistant custom list

Use only model slugs and assistant names returned by the account. When the model catalog is empty, use the account default instead of guessing a slug. Kagi Quick and Kagi Research are usable through --assistant only when they appear in the assistant listing.

Use Assistant deliberately

Use a closed model call for reasoning that needs no current information:

kagi assistant \
  "find the weakest assumption in this design and propose a simpler one" \
  --no-web-access \
  --format markdown

Enable web access when the answer needs current sources:

kagi assistant \
  "compare these database options for this workload and cite current limits" \
  --web-access \
  --format markdown

Choose a live model explicitly only when its cost, speed, or capability matters:

kagi assistant \
  "review this migration plan for data-loss risks" \
  --model MODEL_SLUG \
  --no-web-access \
  --format markdown

For a model comparison, send the same bounded prompt to each model in a fresh thread. Compare correctness, latency, and usefulness, not prose style alone.

Analyze files

Attach only the files needed for the answer. Assistant accepts documents, spreadsheets, images, and audio; one upload can be at most 30 MB.

kagi assistant \
  --attach ./report.pdf \
  "extract the decision, evidence, risks, and unresolved questions" \
  --no-web-access \
  --format markdown

Uploaded content remains in the thread context. Start a new thread when the next task should not inherit it.

Require structured output

Use a built-in contract for reusable decisions, plans, and checklists:

kagi assistant \
  "decide whether this service is ready to launch" \
  --contract decision \
  --format json

Use --contract-file PATH when the consumer needs a different small JSON shape. Contract mode validates the response and makes one repair attempt.

Budget discipline

  1. Use subscriber Summarizer or Translate for direct transformations when the active plan includes them.
  2. Use the account default or a lower-cost model for simple work.
  3. Disable web access for closed-context work.
  4. Keep prompts specific and attach only relevant files.
  5. Start a new thread for an unrelated task.
  6. Reserve Research assistants and premium models for work that needs their extra tools or reasoning depth.

Completion criteria

AI work is complete when:

  • the selected lane matches the task;
  • every explicit model or assistant came from live account discovery;
  • web access and attachments are limited to what the answer needs;
  • structured output passes its contract when a machine will consume it;
  • the response answers the task instead of spending allowance on avoidable transformations; and
  • you changed or deleted account state only when the user asked.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.