Lintel on GitHub Copilot. Generated from
skills/uniformity/SKILL.md; edit the canonical file, then runli-copilot init.
- Resource root:
../../..from this skill's base directory (the Lintel source withbin/,lib/,skills/). Write plans, state and evidence into the working repository's.claude/tree, never into the resource root.- Skill-relative paths: this skill's own
scripts/,references/anddata/folders (and a<base>that the workflow defines as its own directory) mean../../../skills/uniformity/in the Lintel source, not this generated folder.${LINTEL_SKILLS_DIR:-skills}means the skills root,../../../skills. Abin/li-runstep runs in the working repository, so use$LINTEL_SKILLS_DIR/uniformity/there.- Shell steps: run Bash snippets with Bash (Git for Windows'
bash.exeon Windows, neverSystem32\bash.exe). Save a snippet to a temporary.shfile and runbash "<resource root>/bin/li-run" <file>; it preparesLINTEL_SOURCE_ROOT,LINTEL_REPO_ROOTand the profile context.- Tools: Read=
view, Write=create, Edit=edit, Bash=bash/powershell, Grep=grep, Glob=glob, AskUserQuestion=ask_user, TodoWrite=the plan checklist, Task or a named role=taskwith that custom agent, WebFetch=web_fetch.- Other Lintel workflows are native skills: invoke
/li-<name>rather than reading their files. Named roles such asCodeReviewerare custom agents.
You are the uniformity skill — read-only window onto the uniformity-as-contract
mechanism (the parked /autoplan recommendation, now built).
This is contributor-source-only, not consumer health inspection. Use doctor's inspection owner for installed files and target layout. A consumer bundle without the trusted source tests cannot run this floor; retain the visible limitation and never use target-code fallback.
What this skill does
Lintel's 14 uniformity dimensions (D1–D14) are held to a per-kind floor, not a global maximum. This skill surfaces the contract's two enforcement surfaces in one place:
- runs
tests/shape/uniformity-coverage.sh— the Gate-M3 floor check (HARD-FAILS only when aworkflow_root: trueskill or a block-hook lacksnecessity:), which also prints per-kind adoption % of every tracked dimension; - points at
.claude/engineering/audits/uniformity-matrix.md— the living dashboard regenerated bybin/li-uniformity.
It never writes. To regenerate the matrix, the operator runs bin/li-uniformity
explicitly (it edits a doc, so it is not auto-run here).
When to use
- "Is the uniformity floor green?" — runs the shape-test, shows pass/fail
- "What's our necessity-field adoption per kind?" — surfaces the adoption summary
- After adding a new workflow_root skill or block-hook — confirm the floor still holds
- Before a meta-infra REVIEW (Gate M3) — preview the floor verdict
When NOT to use
- To REGENERATE the matrix — that's
bin/li-uniformity(writes a doc) - To CHANGE the contract — edit
docs/concepts/uniformity-contract.md - To enforce the long tail — adoption above the floor is tracked, not mandated; this skill reports it, it does not gate on it
Inputs
All optional:
--check— also runbin/li-uniformity --checkto report whether the matrix is stale (warn-only; does not fail the skill)--matrix— print the matrix's adoption-summary head instead of running the test
Workflow
set -euo pipefail
source_root="${LINTEL_SOURCE_ROOT:?select the trusted Lintel source}"
mode="test"
while [ $# -gt 0 ]; do
case "$1" in
--check) mode="check"; shift ;;
--matrix) mode="matrix"; shift ;;
*) echo "Usage: /li-uniformity [--check] [--matrix]"; exit 2 ;;
esac
done
MATRIX="$source_root/.claude/engineering/audits/uniformity-matrix.md"
floor="$source_root/tests/shape/uniformity-coverage.sh"
generator="$source_root/bin/li-uniformity"
if [ "$mode" = matrix ]; then
[ -f "$MATRIX" ] && [ -r "$MATRIX" ] || {
printf 'UNAVAILABLE: readable source matrix missing: %s\n' "$MATRIX" >&2
exit 2
}
awk '/^## Matrix by kind/{exit} {print}' "$MATRIX"
printf '\nRecorded source matrix only; floor not run. Full matrix: %s\n' "$MATRIX"
exit 0
fi
[ -f "$floor" ] && [ -r "$floor" ] || {
printf 'UNAVAILABLE: source floor helper missing: %s\n' "$floor" >&2
exit 2
}
if bash "$floor"; then :
else
result=$?
printf 'FAILED: floor helper exited %s; preserve its diagnostics above.\n' "$result" >&2
exit "$result"
fi
if [ "$mode" = check ]; then
[ -f "$MATRIX" ] && [ -r "$MATRIX" ] && [ -f "$generator" ] || {
printf 'UNAVAILABLE: source matrix or freshness helper is missing.\n' >&2
exit 2
}
printf '\n## Matrix freshness\n'
result=0
output=$(bash "$generator" --check) || result=$?
printf '%s\n' "$output"
if [ "$result" -ne 0 ]; then
stale='li-uniformity --check: DIFF (matrix is stale; run bin/li-uniformity)'
if [ "$result" -eq 1 ] && { [ "$output" = "$stale" ] || [[ "$output" == "$stale"$'\n'* ]]; }; then
printf 'STALE: source matrix differs; warning only, no regeneration performed.\n'
else
printf 'ERROR: freshness helper exited %s; this is not a staleness verdict.\n' "$result" >&2
exit "$result"
fi
fi
fi
printf '\nSource dashboard: %s\nContract: docs/concepts/uniformity-contract.md\n' "$MATRIX"
Resolve LINTEL_SOURCE_ROOT from the loaded trusted adapter, never from the target cwd.
These helpers inspect their own source component tree; they do not audit an arbitrary
consumer project. A portable bundle may not contain the developer floor test or recorded
matrix: report that boundary, not an invented passing gate or a target-executable fallback.
For an explicitly requested current render,
bash "$LINTEL_SOURCE_ROOT/bin/li-uniformity" --stdout is the existing nonwriting operation.
Do not use default generation or --output
in this read-only workflow.
Status protocol
- PASS — the actual floor helper passed; adoption remains descriptive, not maturity.
- FAILED — preserve the floor helper's nonzero exit and diagnostics. A floor violation names the component; an execution error is not proof that the component violates policy.
- STALE — only the documented diff result is warn-only; no matrix was written.
- UNAVAILABLE / ERROR — a missing resource, unreadable matrix or helper failure leaves its check unverified. Never relabel every nonzero exit as stale.
- RECORDED MATRIX —
--matrixdisplayed existing data; it did not run the floor.
Examples: --check with a missing necessity field retains the floor failure even if the
matrix is also stale. A valid floor and an actual matrix diff return a visible warning.
A missing helper returns unavailable without searching the target or personal install.
None of these outcomes proves that a host executed a declared skill or hook.
Integration
Reads:
tests/shape/uniformity-coverage.sh(the floor gate + adoption reporter).claude/engineering/audits/uniformity-matrix.md(the living dashboard)- every component's frontmatter (indirectly, via the test/generator)
Writes:
- nothing (pure read; matrix regeneration is a separate explicit
bin/li-uniformitycall)
Anti-patterns
- Treating low long-tail adoption as a failure — only the floor (necessity on workflow_root skills + block-hooks) gates. The rest is tracked.
- Hand-editing the matrix — it's generated; edits are overwritten. Change the
source frontmatter and re-run
bin/li-uniformity.
See also
docs/concepts/uniformity-contract.md(the per-kind floor + tracked dimensions)tests/shape/uniformity-coverage.sh(Gate-M3 floor enforcer)bin/li-uniformity(the regenerable matrix generator).claude/engineering/audits/uniformity-matrix.md(the living dashboard)/li-audit(the sibling read-only window onto the unified audit trail).claude/engineering/audits/lintel-uniformity-MASTER.md(the original top-20 findings)