Skip to content

I0815 — Say what is now true in the claims, the documentation and the worker instructions

Add the behavioural claims this milestone earns to docs/CLAIMS.yaml with real implementation and test paths, document the commands in docs/CLI.md, correct the sentence in docs/CONTINUITY.md that says no index is maintained, write docs/KNOWLEDGE.md, and extend the provider-neutral instruction source so a fresh worker knows the session and knowledge steps of the lifecycle.

BLOCKED wave 8 · p1 · deep-work profile · runs alone

Part of M003 — A session is a durable envelope, and knowledge is compiled from what is already canonical.

Blocked. This issue cannot start until I0813, I0814 are done. The status is derived from that, not declared.

Objective

Add the behavioural claims this milestone earns to docs/CLAIMS.yaml with real implementation and test paths, document the commands in docs/CLI.md, correct the sentence in docs/CONTINUITY.md that says no index is maintained, write docs/KNOWLEDGE.md, and extend the provider-neutral instruction source so a fresh worker knows the session and knowledge steps of the lifecycle.

Why

A capability with no claim is invisible, a claim with no test is the failure this project measures in other people's repositories, and a document that contradicts the code teaches the wrong thing to every worker that reads it.

Current state

docs/CONTINUITY.md states that search maintains no index because the corpus is a handful of files, which describes the record corpus correctly and the repository corpus not at all. The instruction source describes the task lifecycle with no session step.

Desired state

The distinction is stated: record search is a direct literal scan with no index, knowledge search is a derived index over a broader corpus, and neither outranks git or the canonical files. Claims name behaviour rather than subcommands. The worker instructions describe the loop from session start to session close, and teach retrieval of what is needed rather than reading the whole base.

Scope

  • docs
  • share/skeleton/providers
  • .majordomus/providers
  • CLAUDE.md
  • AGENTS.md
  • test/cases/73_claims_and_docs.sh

Out of scope

  • Any website change
  • A claim for a subcommand that has no behaviour of its own

Dependencies

What waits on this

Acceptance criteria

  • Every added claim has a source, an implementation and a test path that all exist, and no guaranteed claim lacks a test
  • docs/CONTINUITY.md distinguishes the two searches, and the old sentence no longer contradicts the code
  • The Clean Extraction Boundary and the continuity document agree with each other about the graph
  • The regenerated provider projections stay under the always-loaded budget
  • The instruction source teaches retrieval, not reading the whole base

Validation

  • bin/majordomus doctor
  • bin/majordomus update
  • bash test/run.sh

Evidence required

  • claims_added
  • docs_consistent

Evidence

None recorded. Every token above needs a command or an artifact behind it before this issue can be completed; narrative is refused.

Risk

The always-loaded budget is a hard cap with a failing check. Adding lifecycle steps to the instruction body can push it over, and the fix must be to say less rather than to raise the cap.

Timeline

started
—
verified
—
completed
—

Those three fields, the evidence above and the state of the dependencies are all the status is made of. There is no status field to disagree with them.

Canonical record: .ai/repo/project/issues/I0815.yaml. Read it back with majordomus plan show I0815.