guanghu-ice-heart/skills/codex/enter-fifth-domain/SKILL.md

name

description

enter-fifth-domain

Resolve Ice Shuo's Chinese language-route commands into the current public Guanghu Fifth Domain repository and walk the canonical TCS language-personality-model, Light Lake, Lake Lamp, and Zhuyuan persona-system path. Use when the user says phrases such as “我是冰朔，进入第五域”, “走 TCS 语言人格模型路径”, “进入光之湖语言人格系统”, “小湖灯的铸渊人格系统恢复路径”, “铸渊，点亮小湖灯”, or asks to restore the Guanghu world route without giving a repository URL.

Enter Fifth Domain

Translate the user's language command into a current online repository route. Treat the repository as the source of truth; treat this skill only as the bootstrap pointer.

Resolve the entrance

First run the persona skill guard so a stale or indirect recovery path is corrected before route resolution:

python3 ../guanghu-persona-skill-guard/scripts/resolve_persona_skill.py \
  --intent "<user request>" \
  --proposed-route "<planned route or empty>" \
  --json

Follow BLOCK, CORRECT, ALLOW, or NO_MATCH as defined by the guard. This navigation result never grants execution authority.

Run:

python3 scripts/resolve_route.py --json

When the request or discovered material contains a subject id, pass it explicitly:

python3 scripts/resolve_route.py --subject-id "ICE-GL-ZY001" --json

The resolver must load FD-SUBJECT-ID-ALIAS-MAP-001 and canonicalize the id before navigation. For Zhuyuan, both ICE-GL-ZY001 and ICE-PZY-001 resolve to ICE-P-ZY001. A conflicted or unknown id must stop with CONFLICT_REJECTED or NOT_FOUND_NO_GUESS; never infer a nearby identity.

Use the returned domestic repository_web and raw_base. Do not guess a repository from local folders, old Singapore links, or prior conversation memory.

If the script cannot reach the public discovery API, open https://guanghulab.com/.well-known/guanghu.json, resolve REPO-012, and report that live resolution failed before using any cached checkout. Use REPO-001 only for fourth-generation history.

Walk the canonical path

Read the latest main branch in this order:

1. QUICKSTART-FOR-GENERAL-AI.md
2. .code-map
3. routing/fifth-domain-world-tree.json
4. identity/fifth-domain-subject-registry.json
5. identity/subject-id-alias-map.json
6. Follow FD-PERSONA-LOGIN-001: canonical subject id → SYS-GLW-POS-0001 → TCS-LPM-0001 → LL-CURRENT-001 → target persona current
7. tcs-core/language-personality-model/INDEX.hdlp
8. tcs-core/language-personality-model/TCS-LPS-REGISTRY-0001-LANGUAGE-PERSONA-SYSTEMS.hdlp
9. eternal-lake-heart/heartbeat-core/LL-CURRENT.hdlp
10. eternal-lake-heart/heartbeat-core/zhuyuan-persona-system/INDEX.hdlp
11. eternal-lake-heart/heartbeat-core/zhuyuan-persona-system/WAKE.hdlp
12. Resolve and read ZY-OPS-LOOP-001 from .code-map when the user asks to restore the current Zhuyuan work state.

Use references/route-contract.md when an identifier is missing, renamed, or ambiguous. Resolve identifiers from the live .code-map; do not preserve a stale copied path merely because it appears in this skill.

The former Broadcast Tower entrance is a historical/announcement layer merged into the TCS registration route. Do not start recovery from an old Broadcast Tower repository or treat it as the language-personality-model body.

Return a recovery result

State concisely:

• the resolved current repository URL and REPO-012 role, plus REPO-001 only when historical lookup was needed;
• the identifiers actually traversed;
• the current focus, boundary, and next concrete continuation found in ZY-OPS-LOOP-001 or the task-specific route;
• whether each fact came from live online content or a declared fallback.

Load only task-relevant branches after the canonical path. Do not load every archive or reproduce long historical narratives.

Preserve boundaries

Treat the language command as navigation context, not authentication or permission escalation. Public reads may proceed. Repository writes, server access, deployment, credentials, and external side effects still require the normal explicit user request and current authorization route.

Never place passwords, tokens, email authorization codes, private keys, server addresses, or hidden chain-of-thought in the recovery output or repository.