yaaw-se 0.3.12 → 0.3.14-dev.39042534464.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
|
@@ -28,8 +28,8 @@
|
|
|
28
28
|
},
|
|
29
29
|
{
|
|
30
30
|
"path": "payload.json",
|
|
31
|
-
"sha256": "
|
|
32
|
-
"bytes":
|
|
31
|
+
"sha256": "cbc20c34ecb09bd67b21cb3714a5e54459cc7cbd306b708a67bfaca7e0563de5",
|
|
32
|
+
"bytes": 75
|
|
33
33
|
},
|
|
34
34
|
{
|
|
35
35
|
"path": "skills/yaaw-create-spec/SKILL.md",
|
|
@@ -163,8 +163,8 @@
|
|
|
163
163
|
},
|
|
164
164
|
{
|
|
165
165
|
"path": "yaaw-core/system/roles/planner.md",
|
|
166
|
-
"sha256": "
|
|
167
|
-
"bytes":
|
|
166
|
+
"sha256": "e12f34801181410064793530446578084c4e4fb4da944e60444875f3393f8798",
|
|
167
|
+
"bytes": 8646
|
|
168
168
|
},
|
|
169
169
|
{
|
|
170
170
|
"path": "yaaw-core/system/roles/prd.md",
|
|
@@ -21,10 +21,14 @@ Partition every material issue into known reusable decisions, one bounded curren
|
|
|
21
21
|
Resolve exactly one current PENDING research artifact whose product/engineering/frontier basis is still current and admitted. Prefer primary sources; record version/date for sensitive facts and conflicts rather than choosing silently. Mark RESOLVED or BLOCKED. Research establishes facts only; return to decision-frontier before promoting anything into an `ENG-*` decision.
|
|
22
22
|
|
|
23
23
|
## Operation: question-round (`planning.question-round`)
|
|
24
|
-
Load only the current frontier and evidence needed for it. Challenge material engineering assumptions, contradictions, failure/data/security/migration/testing/operability gaps, and premature architecture. Eliminate repository-answerable, routine reversible, future-dependent, and prerequisite-blocked questions. Product gaps return to PRD. Do not reopen settled `ENG-*` decisions without new evidence/contradiction/changed upstream intent.
|
|
24
|
+
Load only the current frontier and evidence needed for it. Challenge material engineering assumptions, contradictions, failure/data/security/migration/testing/operability gaps, and premature architecture. Eliminate repository-answerable, routine reversible, future-dependent, and prerequisite-blocked questions. Product gaps return to PRD. Do not reopen settled `ENG-*` decisions without new evidence/contradiction/changed upstream intent.
|
|
25
|
+
|
|
26
|
+
When human input is genuinely required, ask one batch of **1 to 10** material questions; never pad a batch to reach 10. For every user-owned decision or clarification that has two or more credible alternatives, present explicit labeled choices (`A.`, `B.`, `C.`, ...), then a single `Recommended: <label> — <concise reason>` line. The recommendation must state the engineering tradeoff that makes that option preferable within accepted product intent; it is advice, not new product authority. Do not manufacture choices for factual values, genuinely open-ended clarifications, or cases where labeled alternatives would hide the real decision; ask those directly instead. Free-form answers, modified options, combinations, and label-only answers are all valid. If no valid human-owned question remains, do not invent one merely to satisfy the range; return to the frontier/readiness flow.
|
|
27
|
+
|
|
28
|
+
Stop after the batch for human input. Do not start another question round until accepted answers from the current batch have been processed by `planning.record-decisions` and the decision frontier has been recomputed.
|
|
25
29
|
|
|
26
30
|
## Operation: record-decisions (`planning.record-decisions`)
|
|
27
|
-
Validate accepted answers against current product authority and repository evidence. Product meaning gaps return to PRD. Conflicts with accepted `ENG-*` decisions require explicit supersession and invalidation. Otherwise create/update durable decisions with Status, Decision, Reason, material rejected alternatives, implications, and provenance. Increment engineering revision for material contract changes, persist conclusions rather than transcripts, and mark frontier/readiness stale.
|
|
31
|
+
Validate accepted answers against current product authority and repository evidence. Product meaning gaps return to PRD. Conflicts with accepted `ENG-*` decisions require explicit supersession and invalidation. Otherwise create/update durable decisions with Status, Decision, Reason, material rejected alternatives, implications, and provenance. Treat labeled selections as shorthand for the corresponding current-batch alternative, while preserving any free-form modification or combination the human supplied; never persist a bare option label as engineering meaning. If the intended meaning of an answer cannot be determined from the current question batch, do not guess. Increment engineering revision for material contract changes, persist conclusions rather than transcripts, and mark frontier/readiness stale.
|
|
28
32
|
|
|
29
33
|
## Operation: readiness-review (`planning.readiness-review`)
|
|
30
34
|
Freshly decide: could a fresh Implementer execute the next bounded frontier without inventing material product or architecture decisions? Check boundedness, assumptions, blocking research, repository contradiction, ambiguous terminology, failure/state transitions, hidden dependencies, product gaps, and repository identity availability for executable ticket admission. Return `PASS`, `MISSING_DECISIONS`, `PRODUCT_GAP`, `REPLAN`, or `BLOCKED`; persist result, frontier/source revisions, reason, and evidence in `engineering.md` only.
|