| name | pretty-shiki-spec |
| description | Generate a Shiki-specific L0 human-friendly HTML spec from L1 consensus specs. |
pretty-shiki-spec
Use this skill when the user asks for a prettier Shiki spec site, a human
review view of Shiki specs, or an L0 spec generated from L1 consensus specs.
Workflow
- Identify the Shiki input path. If no path is given and
shiki_context/
exists, use it.
- Use
--feature <id> when the user wants one feature workspace only.
- Run
tools-skills/skills/pretty-shiki-spec/scripts/publish_pretty_spec.py
with a clear title.
- Use
--fail-on-broken-links when the output is meant for review.
- Report the generated output directory, missing L1 targets, and broken links.
Shiki Layers
- L0 is the generated human-friendly review projection.
- L1 is the consensus spec layer:
_plan.md, index.md, leaf specs,
tests/test_cases.md, and code_contract.md.
- L2 is standards and loadouts such as workflows and tech contracts. L2 should
be cited through L1 references instead of duplicated into L0.
Rules
- Do not modify source L1 Markdown while publishing.
- If L0 and L1 disagree, trust L1 and regenerate after fixing L1.
- Keep generated assets under the requested output directory.
- Do not treat L0 as a coding contract. Coder agents still follow L1 specs.