Each observation an episode recorded becomes, at that episode's end and with no model, a candidate of its own that says what it is about, and knowledge candidates groups the records about one subject with the episodes behind them; an episode that observed nothing records that it found nothing

When an episode ends, every observation it recorded with knowledge observe becomes a knowledge candidate — an observed lesson awaiting review — that says what it is about, without a model and without anyone asking. knowledge candidates groups the records about one subject, so the same defect seen by three sessions reads as one group with three episodes behind it. An episode that observed nothing writes no candidate and still records knowledge.derived with written: 0: it was evaluated and nothing was found.

declared statusdeclared status: guaranteed Declared deterministic and blocking, implemented, and naming the behavioural test that is to prove it. Whether a recorded run supports it is the evidence verdict, shown beside it.

recorded evidencerecorded evidence: not run Declared guaranteed, and not supported by a recorded run: the test is named, and no passing run of the current inputs is recorded. how evidence is recorded

The id is ADR 0091's, the episode followed by a digest of the kind, the subject and the statement, so every file is one episode's own and nothing is read, merged and written back. A rejected record stays rejected and a later observation of its subject is a new candidate listed beside it in the same group. The Rust reader knowledge_base.candidates answers the same groups. ADR 0118.

What it means

When an episode ends, every observation it recorded with knowledge observe becomes a knowledge candidate — an observed lesson awaiting review — that says what it is about, without a model and without anyone asking. knowledge candidates groups the records about one subject, so the same defect seen by three sessions reads as one group with three episodes behind it. An episode that observed nothing writes no candidate and still records knowledge.derived with written: 0: it was evaluated and nothing was found.

How it works

The deriver's table gains one row (ADR 0118): observation.recorded yields a lesson with epistemics observed, about set to the subject and route derived from it. Its id is ADR 0091's — the episode followed by a digest of the kind, the subject and the statement — so every file is one episode's own, a second derivation over the same ledger writes nothing, and two branches that observed the same thing merge without a conflict. Evidence references that resolve now (a commit git knows, a tracked file or test) join derived_from; every reference is listed under # Evidence. The grouping is read, never written: the shell listing and knowledge_base.candidates group candidates and rejected records by about, listing each episode once.

How to see it

majordomus session start
majordomus knowledge observe --kind defect --subject lib/a.sh "a.sh prints the wrong word"
majordomus session end                      # the boundary derives the candidate
majordomus knowledge candidates             # ... about lib/a.sh  route project  awaiting 1  rejected 0  episodes 1
majordomus knowledge candidates --json | jq '.groups'

What it does not cover

It does not decide whether an observation is right, or act on it: a candidate is reviewed, promoted or rejected by a person or an authorised agent. Friction the ledger records on its own (task.refused, task.gate) is not derived; which of those lines are friction rather than the tool working is a later decision.

Why it exists

.ai/repo/adrs/0118-a-session-leaves-what-it-observed-and-one-candidate-per-defect-is-routed.md: sessions should improve the tool by being used, not by being told to reflect, and a recurring defect must read as one thing without any episode writing another's file.

Detail rendered from docs/claims/an-observation-becomes-a-grouped-candidate.md.

From the statement to the proof

1 · the public statement

Each observation an episode recorded becomes, at that episode's end and with no model, a candidate of its own that says what it is about, and knowledge candidates groups the records about one subject with the episodes behind them; an episode that observed nothing records that it found nothing

Made on Layer.
2 · the canonical claim

an-observation-becomes-a-grouped-candidatedeclared status: guaranteed in docs/CLAIMS.yaml

Defined in read it on this site · docs/KNOWLEDGE.md
3 · the implementation

lib/knowledge.sh · majordomus knowledge
4 · the test and the scenarios
Test: test/cases/981_an_observation_becomes_a_grouped_candidate.sh
5 · the recorded runs

recorded evidence: not run how evidence is recorded

Verify it yourself

The test runs in a disposable temporary repository and asserts the behaviour, not a string in the source.

from a clone of the repository

bash test/run.sh 981_an_observation_becomes_a_grouped_candidate

Where this claim is used

responsibilities it covers

Layer

Each observation an episode recorded becomes, at that episode's end and with no model, a candidate of its own that says what it is about, and knowledge candidates groups the records about one subject with the episodes behind them; an episode that observed nothing records that it found nothing

From the statement to the proof

Verify it yourself

Where this claim is used

responsibilities it covers

Related claims same implementation