How do I install the Horosa Agent skill?

Run npx skillmds@latest add horace-maxwell/horosa-agent-2 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Horosa Agent safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Horosa Agent?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Horosa Agent free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under AGPL-3.

Who published Horosa Agent?

horace-maxwell (@horace-maxwell) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Horosa Agent by horace-maxwell — Agent Skill

Name: Horosa Agent
Author: horace-maxwell

Horosa Agent

Use Horosa (星阙) local metaphysics MCP tools correctly — 106 techniques: Western astrology natal chart, solar return, progressions, primary directions, horary, electional 择日, birth-time rectification, 八字 bazi, 紫微斗数 ziwei, 大六壬, 奇门遁甲 qimen, 太乙, 金口诀, 六爻, 梅花心易, 河洛理数, 铁板神数, 塔罗 tarot, 天文地占 geomancy, 老黄历, AstroCartoGraphy, celebrity birth data, Chinese esoteric history. Trigger on 起盘排盘起课起卦算命推运看盘合盘卜卦择日生时校正, fortune-telling, divination, chart casting, or any Horosa/星阙 reading. Never hand-calculate.

File contents

Horosa Agent Skill (Codex / agentskills.io entry)

Thin client-agnostic contract. The single policy source is skills/horosa-agent/SKILL.md — read it (and its references/) before deep work. Maintainer/repo law: AGENTS.md.

Non-negotiable rules

Never hand-calculate any Horosa/星阙 technique (no shell/Python/JS/web formulas, no memorized tables). Every 盘/课/卦/运 comes from a horosa MCP tool call (horosa_astro_chart, horosa_cn_qimen, horosa_cn_liureng_gods, …) or the horosa-skill CLI.
Clarify before calling: if a result-changing setting is missing (birth/event time, place, timezone, gender, house system, 起局方式, 贵人体系, target year), ask the user a short question with options first. Tools reject unconfirmed calls with agent_guidance.required; after the user answers, retry with agent_confirmed_settings: true (or defaults_accepted: true when they accept defaults) plus clarification_notes. Never self-confirm; never switch tools to bypass the gate. In non-interactive runs (codex exec), surface the question as your final answer instead of guessing.
Read results only from export_snapshot.export_text / .sections — that text is the chart. Do not restate numbers that are not in it, and do not invent sections it lacks.
If horosa tools are absent from the tool list: first turn after a cold start may simply be too early (Codex waits ~1s for optional servers; the runtime warms in ~45s) — retry next turn. Still absent → the server is not configured; point the user at examples/clients/codex.md.
Errors carry structured recovery: on details.agent_recovery, relay prompt_to_user verbatim and stop. On transport errors, suggest horosa-skill doctor / horosa-skill selfcheck.

Shell-only agents (no MCP)

codex exec without MCP, CI bots, plain shell tools: the CLI carries the same contract. Parse stdout only (exactly one JSON document; progress lines and error envelopes go to stderr). On Windows PowerShell 5.1 pass payloads as --input / --output files, never through the pipe (it re-encodes bytes and mangles Chinese).

Need	Command
Tool names + `aka:` aliases	`horosa-skill tool list`
What to confirm before a call	`horosa-skill agent guidance --tool qimen`
Run one technique	`horosa-skill tool run qimen --input payload.json --output result.json`
Natural-language routing	`horosa-skill dispatch --input query.json --output result.json`
One-command onboarding on a fresh machine	`horosa-skill setup --client codex` (`uv run horosa-skill setup --client codex` inside a checkout)

Exit 2 + stderr code: "agent_guidance.required" = ask the user with details.agent_recovery.prompt_to_user, then rerun with agent_confirmed_settings: true + clarification_notes (or defaults_accepted: true only when the user explicitly accepts defaults). Exit 0 with ok: false = a real engine error (error.code), not a missing technique. Read results only from data.export_snapshot.export_text; quote data.technique_card afterwards.

---name: horosa-agent-2description: Use Horosa (星阙) local metaphysics MCP tools correctly — 106 techniques: Western astrology natal chart, solar return, progressions, primary directions, horary, electional 择日, birth-time rectification, 八字 bazi, 紫微斗数 ziwei, 大六壬, 奇门遁甲 qimen, 太乙, 金口诀, 六爻, 梅花心易, 河洛理数, 铁板神数, 塔罗 tarot, 天文地占 geomancy, 老黄历, AstroCartoGraphy, celebrity birth data, Chinese esoteric history. Trigger on 起盘排盘起课起卦算命推运看盘合盘卜卦择日生时校正, fortune-telling, divination, chart casting, or any Horosa/星阙 reading. Never hand-calculate.license: AGPL-3.0-only---# Horosa Agent Skill (Codex / agentskills.io entry)Thin client-agnostic contract. The **single policy source** is[`skills/horosa-agent/SKILL.md`](../../../horosa-skill/skills/horosa-agent/SKILL.md) — read it(and its `references/`) before deep work. Maintainer/repo law: `AGENTS.md`.## Non-negotiable rules1. **Never hand-calculate** any Horosa/星阙 technique (no shell/Python/JS/web formulas, no memorized tables). Every 盘/课/卦/运 comes from a `horosa` MCP tool call (`horosa_astro_chart`, `horosa_cn_qimen`, `horosa_cn_liureng_gods`, …) or the `horosa-skill` CLI.2. **Clarify before calling**: if a result-changing setting is missing (birth/event time, place, timezone, gender, house system, 起局方式, 贵人体系, target year), ask the user a short question with options first. Tools reject unconfirmed calls with `agent_guidance.required`; after the user answers, retry with `agent_confirmed_settings: true` (or `defaults_accepted: true` when they accept defaults) plus `clarification_notes`. Never self-confirm; never switch tools to bypass the gate. In non-interactive runs (`codex exec`), surface the question as your final answer instead of guessing.3. **Read results only from `export_snapshot.export_text` / `.sections`** — that text is the chart. Do not restate numbers that are not in it, and do not invent sections it lacks.4. If `horosa` tools are absent from the tool list: first turn after a cold start may simply be too early (Codex waits ~1s for optional servers; the runtime warms in ~45s) — retry next turn. Still absent → the server is not configured; point the user at `examples/clients/codex.md`.5. Errors carry structured recovery: on `details.agent_recovery`, relay `prompt_to_user` verbatim and stop. On transport errors, suggest `horosa-skill doctor` / `horosa-skill selfcheck`.## Shell-only agents (no MCP)`codex exec` without MCP, CI bots, plain shell tools: the CLI carries the same contract. Parse **stdout only**(exactly one JSON document; progress lines and error envelopes go to stderr). On Windows PowerShell 5.1 passpayloads as `--input` / `--output` files, never through the pipe (it re-encodes bytes and mangles Chinese).| Need | Command || --- | --- || Tool names + `aka:` aliases | `horosa-skill tool list` || What to confirm before a call | `horosa-skill agent guidance --tool qimen` || Run one technique | `horosa-skill tool run qimen --input payload.json --output result.json` || Natural-language routing | `horosa-skill dispatch --input query.json --output result.json` || One-command onboarding on a fresh machine | `horosa-skill setup --client codex` (`uv run horosa-skill setup --client codex` inside a checkout) |Exit 2 + stderr `code: "agent_guidance.required"` = ask the user with `details.agent_recovery.prompt_to_user`,then rerun with `agent_confirmed_settings: true` + `clarification_notes` (or `defaults_accepted: true` only whenthe user explicitly accepts defaults). Exit 0 with `ok: false` = a real engine error (`error.code`), not a missingtechnique. Read results only from `data.export_snapshot.export_text`; quote `data.technique_card` afterwards.

Frequently asked questions