agentic-prototype-refinement
This skill is a compatibility shim. It does not own prototype production workflow.
Trigger Conditions
Use this shim when an older entrypoint asks to refine an existing prototype, respond to critique findings, or apply selected-element changes.
For any artifact-producing operation, call agentic-prototype-trigger with profile: refinement plus original source spec, current manifest when present, selected elements when present, and user request or critique diagnostics when present.
Required Inputs
profile: refinement- original source spec when meaning may change
- current manifest when present
selected_elements[]when present- user request when present
- critique diagnostics when present
agentic-prototype-refinement-requestbounding_rectis context only, not a write coordinateuser_instruction
Outputs
- A routing handoff to
agentic-prototype-trigger - Preserved
agentic-designrefinement context for downstream request construction - Blocked questions when selected_elements, user request, or manifest context is incomplete
Forbidden Actions
- This shim must not own prototype production workflow.
- Do not bypass
agentic-prototype-triggerfor any artifact-producing operation. - Do not package or doctor prototype artifacts from this shim.
- Do not commit, push, reset, rebase, stash, or delete user work as part
of this skill (
git_read_onlytier inskills/agentic-git-sync/rules/git-permission-boundaries.md); this does not restrict other active skills' Git contracts.
Command Invocation
Call `agentic-prototype-trigger` with profile: refinement and the collected source spec / manifest / selected_elements / critique context.
Legacy refinement classification text retained for compatibility:
semanticvisualmixed- Classify the request as semantic, visual, or mixed.
- If
user_instructionis missing, ask a read-only clarification and do not write files.
Verification Commands
PYTHONPATH=python .venv/bin/pytest tests/agentic_core/test_design_skills.py -q