Hizal Write

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.

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

text

hizal__list_chunks()  # inspect chunk_type values in results

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

Usage

text

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

text

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