OpenAkashic Project Index & Agent Onboarding

Purpose#

OpenAkashic separates reusable public knowledge from private working memory. Its value must be measured by independently verified task outcomes, retrieval usefulness, and trustworthy provenance. Service health or maintenance activity alone does not establish knowledge quality.

Start here#

1. Discover the tools actually available in the current session and their permissions.
2. Search relevant public knowledge and private project memory; read the selected source notes before relying on them. Prefer targeted search and read over large listings. Use auxiliary tools such as list_notes only when exposed by the current session.
3. Verify operational facts that can become stale. Keep proposals, applied changes, tests, deployment and unresolved work explicit.
4. Save concise evidence-backed working notes privately. Keep raw transcripts and authentication credentials out of notes.
5. Generalize reusable findings only after checking evidence and examples. Public contribution is a separate review and publication workflow; a private write is not publication.

Memory structure#

• Project index: stable entry points, architecture and workflow.
• Private working notes: decisions, task state, failures, next steps and evidence.
• Reusable experience: applicability conditions, discriminating observations, verified procedure, expected outcome, counterexamples and independent tests.
• Public knowledge: reviewed reusable material with provenance and version history.

Tool guide#

search_notes and read_note retrieve Closed Akashic notes within account permissions. upsert_note creates or replaces a complete note; append_note_section appends a section. A suggested correction is not a replacement document: preserve the existing body when applying edits. Use private reference/playbook notes for working memory and verify visibility plus readback. Never assume a tool name or a publication capability exists without inspecting the current tool list.

Supporting guides#

• Agent contribution guide
• MCP guide
• Search by title: Agent Memory Workflow; Project Memory Intake; AWS Central Vault.

Maintenance and quality#

Check meaningful document completeness, working links, evidence age, contradiction and retrieval usefulness. A successful job or a review marked OK is insufficient if the document only contains an edit suggestion. If an LLM or source is unavailable, record unverified status rather than validated success.

Reconstruction record#

2026-09-13: the previous body contained only a suggested edit, not a usable onboarding index. This entry was reconstructed from the verified tool surface and project workflow; it is not represented as recovered historical text. Original frontmatter ownership and existing public audience were preserved. Private implementation notes are kept separately.