Stage 1 checklist and failure modes
Outcome
This page is a single place to work from when running all of Stage 1, from framing the domain to closing the last gap. It does not repeat any sub-stage page’s own conditions or failures; it links to them. Its own added value is two lists this guide keeps nowhere else: every Stage 1 gate in page order, and a short set of failure patterns that show up on more than one sub-stage.
Checklist by sub-stage
Each row links to that sub-stage’s own “Definition of done” and “Common failures” sections. Work through the pages in this order; a later sub-stage assumes the ones above it are done.
| Sub-stage | Definition of done | Common failures |
|---|---|---|
| S1.1 and S1.2, frame the domain and list the tasks | Definition of done | Common failures |
| S1.3, draft the blueprint | Definition of done | Common failures |
| S1.4a, search planning and execution | Definition of done | Common failures |
| S1.4b, screening and conversion | Definition of done | Common failures |
| S1.4c, injection screening | Definition of done | Common failures |
| S1.5a, concept extraction | Definition of done | Common failures |
| S1.5b, distillate and quote bank | Definition of done | Common failures |
| S1.6, knowledge items | Definition of done | Common failures |
| S1.7, concept map and prerequisite hierarchy | Definition of done | Common failures |
| S1.8, coverage and gaps | Definition of done | Common failures |
Gates in order
A trimmed, page-ordered view of the Stage 1 index’s own Approvals table. That table stays the source of truth; read it for the human-roles gate kind and the basis label behind each row.
| After | What the person approves | Where it is defined |
|---|---|---|
| S1.1 | The boundary statement | Approvals |
| S1.3 | The weighted blueprint | Approvals |
| S1.4a | The search plan set, before any query runs | Approvals |
| S1.4a, the runner | What to do when the circuit breaker opens | Approvals |
| S1.4b and S1.4c | The selected sources, the flagged list and the blocked-document list | Approvals |
| S1.5 | Each batch of distillations | Approvals |
| S1.6 | A batch of sources’ draft items | Approvals |
| S1.6 | The consolidated, deduplicated item set | Approvals |
| S1.7 | The concept map and hierarchy | Approvals |
| S1.8 | The expansion plan | Approvals |
Failure patterns that recur
Each pattern below shows up more than once across Stage 1. The link goes to a page where a concrete instance is described in full; the pattern itself is not repeated there.
- A check that passes on a small, synthetic sample but says nothing about a reader’s own material. A clean run of a sample script only shows that the script works on this guide’s own data; see S1.4c’s tripwire note and S1.5b’s quote-check limits for two stated instances.
- A pause satisfied by a standing instruction to continue, which no longer shows that anyone looked. See Human roles, gates and batching for the general point, and the S1.5 batch-pause row for where this guide flags it as a real risk rather than a formality.
- A later step using an artifact before its own approval has actually landed. S1.4a requires the search plan set approved before any query runs; S1.7 similarly holds the map and hierarchy for approval before S1.8 reads it. Skipping ahead defeats the point of the gate.
- A cap or threshold that nothing enforces automatically, so a person has to remember to check it. S1.4a’s own recall target, S1.7’s relation-density band, and S1.8’s expansion-loop cap are each a number this guide’s own scripts do not stop the reader from exceeding; a person has to look. (The circuit breaker on S1.4a’s own runner is the opposite case: it does stop the queue by itself, and only the decision of what to do next needs a person.)
Before you start Stage 2
Confirm each of these exists and has a recorded approval before moving on:
- A dated boundary file and a task list (S1.1 and S1.2).
- A blueprint that passes its check, with a weight rationale for every domain (S1.3).
- Admitted sources: converted, checked, and scanned for hidden instructions (S1.4a through S1.4c).
- Canonical knowledge items, deduplicated across sources (S1.6).
- A checked concept map and prerequisite hierarchy, with every concept’s tier filled in (S1.7).
- A gap list with a decision recorded for every open row: closed, deferred, or accepted as a known gap (S1.8).
Next: Stage 2 Content development.