菜单
Use for loss-bounded context compression into NATASHA context packets with fidelity metrics and NO_OP when compression would drop hard requirements.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | token-web |
| description | Use for loss-bounded context compression into NATASHA context packets with fidelity metrics and NO_OP when compression would drop hard requirements. |
| model_adapter | gemini |
| system_instruction | **Module:** **COMPRESS** · Parent: [`187web-ecosystem`](../187web-ecosystem/SKILL.md) |
Gemini adapter: Load as a system instruction. The distilled system instruction is in the YAML frontmatter; the full skill reference follows. Source:
../../.claude/skills/token-web/SKILL.md.
Module: COMPRESS · Parent: 187web-ecosystem
Phase 1 foundation stub: full references/templates expand in Phase 2. Agents must already obey compression invariants below.
/187 compress, COMPRESS, token-web, context capsule
oversized context, subagent handoff, multi-file packet, token budget
NO_OPRaw mission text, constraints, paths, numbers, decisions, token_budget.
NO_OPZero false additions. Prefer NO_OP over silent loss of requirements.
NO_OP.Versioning, changelogs, releases, migration notes, compatibility, deprecations, and adapter sync.
Use when routing across the full 187SKILLS suite or the NATASHA multi-agent integration stack (THREAD, COMPRESS, TENSION, SPARK, CORD, SCOUT, LAB, FUSE).
Use for EVM-first smart-contract assurance, DeFi economic risk analysis, severity/confidence separation, and responsible disclosure (no live keys or exploits).
Use for quantum algorithm selection, circuit design, optimization, resource estimation, and claim-disciplined audits (no unsupported advantage claims).
Use when generating a new repository, deploying to GitHub, creating an installer site, or routing the 187SKILLS suite.
Use when adding delight, micro-interactions, onboarding, community, or retention to a 187WEB project.