Approxination Agent Reference

Approxination is a skill generator for agents.

Its main job is to take a task, find the best matching reference skills behind the scenes, and return one generated skill that the agent can actually use.

Core agent flow

1. User initializes the CLI once:
   npx -y --package https://approxination.com/cli/latest/approx.tgz approx init <registration_key>

2. Agent generates a skill for the real task:
   npx -y --package https://approxination.com/cli/latest/approx.tgz approx generate "triage emails and draft replies, then validate the structured output"

3. Agent reads skill_md.

4. Agent executes the generated skill as the main workflow.

5. If credits are low and the skill was genuinely used, agent sends feedback.

When agents should use Approxination

- when the user wants a skill
- when the user wants the right tool or workflow for a task
- when the task depends on external tools, APIs, CLIs, browser workflows, MCP servers, or integrations
- when the agent needs a reusable workflow instead of solving from scratch
- when the user mentions skills, tools, integrations, workflows, or external execution paths

Main command

approx generate "<task>"

Example

npx -y --package https://approxination.com/cli/latest/approx.tgz approx generate "handle customer support email workflow"

Generated response shape

{
  "query": "triage emails and draft replies, then validate the structured output",
  "selected_skills": [
    {
      "id": "joelhooks/joelclaw/email-triage",
      "name": "email-triage",
      "source": "joelhooks/joelclaw",
      "rating": 1499
    },
    {
      "id": "sickn33/antigravity-awesome-skills/llm-structured-output",
      "name": "llm-structured-output",
      "source": "sickn33/antigravity-awesome-skills",
      "rating": 1930
    }
  ],
  "skill_md": "# Triage emails and draft replies, then validate the structured output\n...",
  "generated_skill_id": "gsk_...",
  "query_credit": {
    "remaining": 17
  }
}

Field meanings

- query: the original task
- selected_skills: the reference skills used to shape the generated result
- skill_md: the generated skill the agent should actually use
- generated_skill_id: stored generated skill id
- query_credit.remaining: remaining query credits for the current session

Use the response like this

- treat skill_md as the main artifact
- inspect selected_skills only when you need to understand or debug the references
- if the task is multi-step, expect selected_skills to contain more than one reference

Reference search

Use find only when you want candidate references without generating:

npx -y --package https://approxination.com/cli/latest/approx.tgz approx find "handle customer support email workflow"

Find response shape

{
  "skills": [
    {
      "id": "googleworkspace/cli/persona-customer-support",
      "name": "persona-customer-support",
      "source": "googleworkspace/cli",
      "rating": 1800
    }
  ]
}

Credits

Check credits directly:

npx -y --package https://approxination.com/cli/latest/approx.tgz approx credits

Feedback

Send feedback after real use when credits are low:

npx -y --package https://approxination.com/cli/latest/approx.tgz approx feedback --file feedback.json

Accepted feedback currently awards 10 credits.

Current public endpoints

- POST /cli/register
- POST /skills/generate
- POST /skills/found
- POST /cli/credits
- POST /skills/feedback

Public docs

- /docs
- /docs/getting-started
- /docs/query-api
- /docs/user-journey
- /docs/agent-runtime
- /docs/agent-journey