name: ll-dev-feat-to-tech description: Governed LL workflow skill for transforming one frozen FEAT inside a feat_freeze_package into a TECH-first design package with optional ARCH and API companions before downstream tech-impl work.
LL Dev FEAT to TECH
This skill is a lite-native governed workflow between ll-product-epic-to-feat and downstream tech-impl consumers. It turns one selected frozen FEAT into a design package where TECH is mandatory and ARCH / API are conditionally derived companions.
Canonical Authority
- Workflow template:
E:\ai\LEE\spec-global\departments\dev\workflows\templates\tech-design-l3-template.yaml - Upstream handoff:
ll-product-epic-to-feat - Downstream workflow:
workflow.dev.tech_to_impl - Primary runtime command:
python scripts/feat_to_tech.py run --input <feat-package-dir> --feat-ref <feat-ref> --repo-root <repo-root>
Runtime Boundary Baseline
- Classify this capability under
E:\ai\LEE-Lite-skill-first\ssot\adr\ADR-038-运行时核心抽象边界与对象分层基线.MD. - This skill is a governed
SkillandWorkflowauthority forFEAT -> TECHderivation. Its scripts and package files are carriers, not independent runtime authorities.
Required Read Order
ll.contract.yamlinput/contract.yamloutput/contract.yamlresources/upstream-workflow-analysis.mdagents/executor.mdagents/supervisor.mdinput/semantic-checklist.mdandoutput/semantic-checklist.md
Execution Protocol
- Accept only a freeze-ready
feat_freeze_packageemitted byll-product-epic-to-feat, plus an explicitfeat_ref, plus an explicitintegration_context. - Validate the package structurally before drafting or editing any design output.
- Resolve the authoritative FEAT context from
feat-freeze-bundle.md,feat-freeze-bundle.json, inheritedsource_refs, the upstream acceptance evidence, and the frozenintegration_context. - Run
python scripts/feat_to_tech.py executor-run --input <feat-package-dir> --feat-ref <feat-ref>to generate the governed design package. - Always produce
TECH; produceARCHonly when the FEAT is architecture-impacting; produceAPIonly when a cross-boundary contract exists. - Record execution evidence, then hand the package to the supervisor.
- Run
python scripts/feat_to_tech.py supervisor-review --artifacts-dir <tech-package-dir>before freeze. - Freeze only after the supervisor records a semantic pass and
python scripts/feat_to_tech.py freeze-guard --artifacts-dir <tech-package-dir>returns success. - Emit a downstream handoff that preserves
feat_ref,tech_ref, optionalarch_ref/api_ref, and the full integration and ownership refs needed byworkflow.dev.tech_to_impl. - When external gate returns
reviseorretry, rerunrun,executor-run, orsupervisor-reviewwith--revision-request <revision-request.json>so the regenerated design package preserves normalized revision context and evidence.
Workflow Boundary
- Input: one
feat_freeze_packageplus one selectedfeat_refplusintegration_context - Output: one
tech_design_packagecontaining a mandatoryTECHdesign object and optionalARCH/APIdesign companions - Out of scope: direct implementation planning, TASK authoring, TESTSET authoring, or bypassing the governed FEAT to TECH boundary
Non-Negotiable Rules
- Do not accept raw requirements, SRC candidates, EPIC packages, or standalone FEAT markdown outside a governed
feat_freeze_package. - Do not treat
ARCHandAPIas unconditional peers ofTECH; they are conditional child artifacts decided by need assessment. - Do not let
ARCH,TECH, andAPIrestate the same material.ARCHowns system placement and boundaries,TECHowns implementation design,APIowns external contracts. - Do not bypass the final cross-artifact consistency check before freeze.
- Do not let the executor self-approve semantic validity.
- Do not freeze unless
integration_context_sufficientandstateful_design_presentare explicitly recorded. - Do not let
TECHomit internal state machine, key algorithm, input/output side effects, or canonical ownership facts that downstream implementation must inherit.