360 Secure SkillHub
360 安全技能中心

Hizal Write

该 Skill 定义了 Hizal 写入工具集的使用规范:一旦智能体做出架构决策、发现代码库约定、学到经验或踩到坑,就立即调用对应工具写入记忆、知识、约定、身份或组织知识等 chunk,并支持自定义类型与个人记忆提升到项目范围。

parkerscobey
parkerscobey
热度 16.02
最近更新 2026-05-23 02:25:21
分类:知识依赖:MCP 服务应用场景:笔记整理 / 工作流 / 任务管理

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add hizal-write
Metadata
name
hizal-write
description
Persist what the agent learns as it builds. Self-triggering — fires whenever the agent makes a decision, discovers a pattern, learns a convention, hits a gotcha, or gains knowledge worth keeping. Use continuously, not just at the end. Triggers on phrases like "good to know", "I'll remember that", "worth noting", "that's a pattern", "lesson learned", "this is how we do it", or any moment the agent learns something it would otherwise forget.

Hizal Write

Write as you build. Not optional.

Available Write Tools

What you're writing Tool Scope
Personal observation or lesson learned hizal__write_memory AGENT
Architecture or design decision hizal__write_knowledge PROJECT
Convention this codebase follows hizal__write_convention PROJECT (auto-inject)
Agent identity or personality hizal__write_identity AGENT (auto-inject)
Org-wide knowledge hizal__write_org_knowledge ORG
Principle (requires human promotion) hizal__store_principle ORG
Custom chunk type hizal__write_chunk varies

Do not use hizal__write_context — it's deprecated. Use the purpose-built tools above.

Custom Chunk Types via write_chunk

For chunk types not covered by the built-in tools, use hizal__write_chunk. Chunk types are customizable per org — always discover what's available before assuming a type doesn't exist.

Discover Available Types

hizal__list_chunks()  # inspect chunk_type values in results

Or check the project's AGENTS.md for the org's type conventions.

Usage

hizal__write_chunk(
  type="<chunk-type-slug>",
  query_key="<unique-key>",
  title="<short title>",
  content="<full content>",
  scope="PROJECT",  # or AGENT, ORG
  project_id="<id>",  # required for PROJECT scope
  org_id="<id>",  # required for ORG scope
  agent_id="<id>"  # required for AGENT scope
)

write_chunk accepts the same common fields as all other write tools (query_key, title, content, gotchas, related, source_file, inject_audience, etc.). It also accepts custom_fields for type-specific metadata.

When to Use write_chunk vs Built-in Tools

  • Built-in tools — use whenever the content maps cleanly to MEMORY, KNOWLEDGE, CONVENTION, IDENTITY, ORG_KNOWLEDGE, or PRINCIPLE
  • write_chunk — use for any custom org type (e.g., SPEC, RUNBOOK, ADR, INVESTIGATION)
  • When in doubt, check what chunk types the org has defined and pick the closest match

Write One Chunk Per Decision

Don't batch everything into one chunk at the end. Write as you go:

  • Made an architecture decision? → write_knowledge now
  • Learned a codebase convention? → write_convention now
  • Discovered something useful personally? → write_memory now
  • Found a custom type that fits better? → write_chunk now

Common Fields

All write tools accept:

  • query_key — unique key for this topic (enables exact lookup later)
  • title — short descriptive title
  • content — the full context content
  • source_file / source_lines — where this came from
  • gotchas — list of warnings or pitfalls
  • related — list of related query_keys
  • inject_audience — DNF targeting spec for auto-injection (omit for defaults)

Example

hizal__write_knowledge(
  project_id="<id>",
  query_key="nuvei-webhook-signing",
  title="DMN Webhook Signature Verification",
  content="All DMN payloads are verified using SHA-256 HMAC. The checksum is computed from sorted key-value pairs + merchant secret. See package internal/checksum for implementation.",
  related=["nuvei-payload-structure", "merchant-credentials"],
  gotchas=["Never use /payment for credential verification"]
)

Promote AGENT → PROJECT

If a personal memory chunk is broadly useful for the team, promote it:

  1. Read the chunk with hizal__read_context
  2. Write it back as write_knowledge or write_convention with the content
  3. Optionally delete the original AGENT chunk with hizal__delete_context