Add static eval via skill-validator (#1195)

* Add static eval via skill-validator

* Add issues: write permission for PR comment posting
This commit is contained in:
Jan Krivanek
2026-03-30 03:02:11 +02:00
committed by GitHub
parent 819a8fa89c
commit 85d690908b
2 changed files with 537 additions and 0 deletions

217
.github/workflows/skill-check.yml vendored Normal file
View File

@@ -0,0 +1,217 @@
name: Skill Validator — PR Gate
on:
pull_request:
branches: [staged]
types: [opened, synchronize, reopened]
paths:
- "skills/**"
- "agents/**"
- "plugins/**/skills/**"
- "plugins/**/agents/**"
permissions:
contents: read
pull-requests: write
issues: write
jobs:
skill-check:
runs-on: ubuntu-latest
steps:
- name: Checkout code
uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4.3.1
with:
fetch-depth: 0
# ── Download & cache skill-validator ──────────────────────────
- name: Get cache key date
id: cache-date
run: echo "date=$(date +%Y-%m-%d)" >> "$GITHUB_OUTPUT"
- name: Restore skill-validator from cache
id: cache-sv
uses: actions/cache/restore@cdf6c1fa76f9f475f3d7449005a359c84ca0f306 # v5.0.3
with:
path: .skill-validator
key: skill-validator-linux-x64-${{ steps.cache-date.outputs.date }}
restore-keys: |
skill-validator-linux-x64-
- name: Download skill-validator
if: steps.cache-sv.outputs.cache-hit != 'true'
run: |
mkdir -p .skill-validator
curl -fsSL \
"https://github.com/dotnet/skills/releases/download/skill-validator-nightly/skill-validator-linux-x64.tar.gz" \
-o .skill-validator/skill-validator-linux-x64.tar.gz
tar -xzf .skill-validator/skill-validator-linux-x64.tar.gz -C .skill-validator
rm .skill-validator/skill-validator-linux-x64.tar.gz
chmod +x .skill-validator/skill-validator
- name: Save skill-validator to cache
if: steps.cache-sv.outputs.cache-hit != 'true'
uses: actions/cache/save@cdf6c1fa76f9f475f3d7449005a359c84ca0f306 # v5.0.3
with:
path: .skill-validator
key: skill-validator-linux-x64-${{ steps.cache-date.outputs.date }}
# ── Detect changed skills & agents ────────────────────────────
- name: Detect changed skills and agents
id: detect
run: |
CHANGED_FILES=$(git diff --name-only origin/${{ github.base_ref }}...HEAD)
# Extract unique skill directories that were touched
SKILL_DIRS=$(echo "$CHANGED_FILES" | grep -oP '^skills/[^/]+' | sort -u || true)
# Extract agent files that were touched
AGENT_FILES=$(echo "$CHANGED_FILES" | grep -oP '^agents/[^/]+\.agent\.md$' | sort -u || true)
# Extract plugin skill directories
PLUGIN_SKILL_DIRS=$(echo "$CHANGED_FILES" | grep -oP '^plugins/[^/]+/skills/[^/]+' | sort -u || true)
# Extract plugin agent files
PLUGIN_AGENT_FILES=$(echo "$CHANGED_FILES" | grep -oP '^plugins/[^/]+/agents/[^/]+\.agent\.md$' | sort -u || true)
# Build CLI arguments for --skills
SKILL_ARGS=""
for dir in $SKILL_DIRS $PLUGIN_SKILL_DIRS; do
if [ -d "$dir" ]; then
SKILL_ARGS="$SKILL_ARGS $dir"
fi
done
# Build CLI arguments for --agents
AGENT_ARGS=""
for f in $AGENT_FILES $PLUGIN_AGENT_FILES; do
if [ -f "$f" ]; then
AGENT_ARGS="$AGENT_ARGS $f"
fi
done
SKILL_COUNT=$(echo "$SKILL_ARGS" | xargs -n1 2>/dev/null | wc -l || echo 0)
AGENT_COUNT=$(echo "$AGENT_ARGS" | xargs -n1 2>/dev/null | wc -l || echo 0)
TOTAL=$((SKILL_COUNT + AGENT_COUNT))
echo "skill_args=$SKILL_ARGS" >> "$GITHUB_OUTPUT"
echo "agent_args=$AGENT_ARGS" >> "$GITHUB_OUTPUT"
echo "total=$TOTAL" >> "$GITHUB_OUTPUT"
echo "skill_count=$SKILL_COUNT" >> "$GITHUB_OUTPUT"
echo "agent_count=$AGENT_COUNT" >> "$GITHUB_OUTPUT"
echo "Found $SKILL_COUNT skill dir(s) and $AGENT_COUNT agent file(s) to check."
# ── Run skill-validator check ─────────────────────────────────
- name: Run skill-validator check
id: check
if: steps.detect.outputs.total != '0'
run: |
SKILL_ARGS="${{ steps.detect.outputs.skill_args }}"
AGENT_ARGS="${{ steps.detect.outputs.agent_args }}"
CMD=".skill-validator/skill-validator check --verbose"
if [ -n "$SKILL_ARGS" ]; then
CMD="$CMD --skills $SKILL_ARGS"
fi
if [ -n "$AGENT_ARGS" ]; then
CMD="$CMD --agents $AGENT_ARGS"
fi
echo "Running: $CMD"
# Capture output; don't fail the workflow (warn-only mode)
set +e
OUTPUT=$($CMD 2>&1)
EXIT_CODE=$?
set -e
echo "exit_code=$EXIT_CODE" >> "$GITHUB_OUTPUT"
# Save output to file (multi-line safe)
echo "$OUTPUT" > sv-output.txt
echo "$OUTPUT"
# ── Post / update PR comment ──────────────────────────────────
- name: Post PR comment with results
if: steps.detect.outputs.total != '0'
uses: actions/github-script@f28e40c7f34bde8b3046d885e986cb6290c5673b # v7.1.0
with:
script: |
const fs = require('fs');
const marker = '<!-- skill-validator-results -->';
const output = fs.readFileSync('sv-output.txt', 'utf8').trim();
const exitCode = '${{ steps.check.outputs.exit_code }}';
const skillCount = parseInt('${{ steps.detect.outputs.skill_count }}', 10);
const agentCount = parseInt('${{ steps.detect.outputs.agent_count }}', 10);
const totalChecked = skillCount + agentCount;
// Count errors, warnings, advisories from output
const errorCount = (output.match(/\bError\b/gi) || []).length;
const warningCount = (output.match(/\bWarning\b/gi) || []).length;
const advisoryCount = (output.match(/\bAdvisory\b/gi) || []).length;
let statusLine;
if (errorCount > 0) {
statusLine = `**${totalChecked} resource(s) checked** | ⛔ ${errorCount} error(s) | ⚠️ ${warningCount} warning(s) | ${advisoryCount} advisory(ies)`;
} else if (warningCount > 0) {
statusLine = `**${totalChecked} resource(s) checked** | ⚠️ ${warningCount} warning(s) | ${advisoryCount} advisory(ies)`;
} else {
statusLine = `**${totalChecked} resource(s) checked** | ✅ All checks passed`;
}
const body = [
marker,
'## 🔍 Skill Validator Results',
'',
statusLine,
'',
'<details>',
'<summary>Full output</summary>',
'',
'```',
output,
'```',
'',
'</details>',
'',
exitCode !== '0'
? '> **Note:** Errors were found. These are currently reported as warnings and do not block merge. Please review and address when possible.'
: '',
].join('\n');
// Find existing comment with our marker
const { data: comments } = await github.rest.issues.listComments({
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: context.issue.number,
per_page: 100,
});
const existing = comments.find(c => c.body.includes(marker));
if (existing) {
await github.rest.issues.updateComment({
owner: context.repo.owner,
repo: context.repo.repo,
comment_id: existing.id,
body,
});
console.log(`Updated existing comment ${existing.id}`);
} else {
await github.rest.issues.createComment({
owner: context.repo.owner,
repo: context.repo.repo,
issue_number: context.issue.number,
body,
});
console.log('Created new PR comment');
}
- name: Post skip notice if no skills changed
if: steps.detect.outputs.total == '0'
run: echo "No skill or agent files changed in this PR — skipping validation."

View File

@@ -0,0 +1,320 @@
name: Skill Quality Report — Nightly Scan
on:
schedule:
- cron: "0 3 * * *" # 3:00 AM UTC daily
workflow_dispatch: # allow manual trigger
permissions:
contents: read
discussions: write
issues: write # fallback if Discussions are not enabled
jobs:
nightly-scan:
runs-on: ubuntu-latest
steps:
- name: Checkout code
uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4.3.1
with:
fetch-depth: 0 # full history for git-log author fallback
# ── Download & cache skill-validator ──────────────────────────
- name: Get cache key date
id: cache-date
run: echo "date=$(date +%Y-%m-%d)" >> "$GITHUB_OUTPUT"
- name: Restore skill-validator from cache
id: cache-sv
uses: actions/cache/restore@cdf6c1fa76f9f475f3d7449005a359c84ca0f306 # v5.0.3
with:
path: .skill-validator
key: skill-validator-linux-x64-${{ steps.cache-date.outputs.date }}
restore-keys: |
skill-validator-linux-x64-
- name: Download skill-validator
if: steps.cache-sv.outputs.cache-hit != 'true'
run: |
mkdir -p .skill-validator
curl -fsSL \
"https://github.com/dotnet/skills/releases/download/skill-validator-nightly/skill-validator-linux-x64.tar.gz" \
-o .skill-validator/skill-validator-linux-x64.tar.gz
tar -xzf .skill-validator/skill-validator-linux-x64.tar.gz -C .skill-validator
rm .skill-validator/skill-validator-linux-x64.tar.gz
chmod +x .skill-validator/skill-validator
- name: Save skill-validator to cache
if: steps.cache-sv.outputs.cache-hit != 'true'
uses: actions/cache/save@cdf6c1fa76f9f475f3d7449005a359c84ca0f306 # v5.0.3
with:
path: .skill-validator
key: skill-validator-linux-x64-${{ steps.cache-date.outputs.date }}
# ── Run full scan ─────────────────────────────────────────────
- name: Run skill-validator check on all skills
id: check-skills
run: |
set +e
set -o pipefail
.skill-validator/skill-validator check \
--skills ./skills \
--verbose \
2>&1 | tee sv-skills-output.txt
echo "exit_code=${PIPESTATUS[0]}" >> "$GITHUB_OUTPUT"
set +o pipefail
set -e
- name: Run skill-validator check on all agents
id: check-agents
run: |
set +e
set -o pipefail
AGENT_FILES=$(find agents -name '*.agent.md' -type f 2>/dev/null | tr '\n' ' ')
if [ -n "$AGENT_FILES" ]; then
.skill-validator/skill-validator check \
--agents $AGENT_FILES \
--verbose \
2>&1 | tee sv-agents-output.txt
echo "exit_code=${PIPESTATUS[0]}" >> "$GITHUB_OUTPUT"
else
echo "No agent files found."
echo "" > sv-agents-output.txt
echo "exit_code=0" >> "$GITHUB_OUTPUT"
fi
set +o pipefail
set -e
# ── Build report with author attribution ──────────────────────
- name: Build quality report
id: report
uses: actions/github-script@f28e40c7f34bde8b3046d885e986cb6290c5673b # v7.1.0
with:
script: |
const fs = require('fs');
const path = require('path');
const { execSync } = require('child_process');
// ── Parse CODEOWNERS ──────────────────────────────────
function parseCodeowners() {
const map = new Map();
try {
const raw = fs.readFileSync('CODEOWNERS', 'utf8');
for (const line of raw.split('\n')) {
const trimmed = line.trim();
if (!trimmed || trimmed.startsWith('#')) continue;
const parts = trimmed.split(/\s+/);
if (parts.length >= 2) {
const filePath = parts[0].replace(/^\//, '').replace(/\/$/, '');
const owners = parts.slice(1).filter(p => p.startsWith('@'));
if (owners.length > 0) {
map.set(filePath, owners);
}
}
}
} catch (e) {
console.log('Could not parse CODEOWNERS:', e.message);
}
return map;
}
// ── Resolve author for a path ─────────────────────────
function resolveAuthor(resourcePath, codeowners) {
// CODEOWNERS semantics: last matching rule wins.
// Also treat "*" as a match-all default rule.
let matchedOwners = null;
for (const [pattern, owners] of codeowners) {
if (
pattern === '*' ||
resourcePath === pattern ||
resourcePath.startsWith(pattern + '/')
) {
matchedOwners = owners;
}
}
if (matchedOwners && matchedOwners.length > 0) {
return matchedOwners.join(', ');
}
// Fallback: git log
try {
const author = execSync(
`git log --format='%aN' --follow -1 -- "${resourcePath}"`,
{ encoding: 'utf8' }
).trim();
return author || 'unknown';
} catch {
return 'unknown';
}
}
// ── Parse skill-validator output ──────────────────────
// The output is a text report; we preserve it as-is and
// augment it with author info in the summary.
const skillsOutput = fs.readFileSync('sv-skills-output.txt', 'utf8').trim();
const agentsOutput = fs.existsSync('sv-agents-output.txt')
? fs.readFileSync('sv-agents-output.txt', 'utf8').trim()
: '';
const codeowners = parseCodeowners();
// Count findings
const combined = skillsOutput + '\n' + agentsOutput;
const errorCount = (combined.match(/\bError\b/gi) || []).length;
const warningCount = (combined.match(/\bWarning\b/gi) || []).length;
const advisoryCount = (combined.match(/\bAdvisory\b/gi) || []).length;
// Count total skills & agents checked
let skillDirs = [];
try {
skillDirs = fs.readdirSync('skills', { withFileTypes: true })
.filter(d => d.isDirectory())
.map(d => d.name);
} catch {}
let agentFiles = [];
try {
agentFiles = fs.readdirSync('agents')
.filter(f => f.endsWith('.agent.md'));
} catch {}
// ── Build author-attributed summary ───────────────────
// Extract per-resource blocks from output. The validator
// prints skill names as headers — we annotate them with
// the resolved owner.
function annotateWithAuthors(output, kind) {
if (!output) return '_No findings._';
const lines = output.split('\n');
const annotated = [];
for (const line of lines) {
// Skill names appear as headers, e.g. "## skill-name" or "skill-name:"
const headerMatch = line.match(/^(?:#{1,3}\s+)?([a-z0-9][a-z0-9-]+(?:\.[a-z0-9.-]+)?)\b/);
if (headerMatch) {
const name = headerMatch[1];
const resourcePath = kind === 'skill'
? `skills/${name}`
: `agents/${name}.agent.md`;
const author = resolveAuthor(resourcePath, codeowners);
annotated.push(`${line} — ${author}`);
} else {
annotated.push(line);
}
}
return annotated.join('\n');
}
const today = new Date().toISOString().split('T')[0];
const title = `Skill Quality Report — ${today}`;
const body = [
`# ${title}`,
'',
`**${skillDirs.length} skills** and **${agentFiles.length} agents** scanned.`,
'',
`| Severity | Count |`,
`|----------|-------|`,
`| ⛔ Errors | ${errorCount} |`,
`| ⚠️ Warnings | ${warningCount} |`,
`| Advisories | ${advisoryCount} |`,
'',
'---',
'',
'## Skills',
'',
'<details>',
'<summary>Full skill-validator output for skills</summary>',
'',
'```',
annotateWithAuthors(skillsOutput, 'skill'),
'```',
'',
'</details>',
'',
'## Agents',
'',
'<details>',
'<summary>Full skill-validator output for agents</summary>',
'',
'```',
annotateWithAuthors(agentsOutput, 'agent'),
'```',
'',
'</details>',
'',
'---',
'',
`_Generated by the [Skill Validator nightly scan](https://github.com/${context.repo.owner}/${context.repo.repo}/actions/workflows/skill-quality-report.yml)._`,
].join('\n');
core.setOutput('title', title);
core.setOutput('body_file', 'report-body.md');
fs.writeFileSync('report-body.md', body);
# ── Create Discussion (preferred) or Issue (fallback) ────────
- name: Create Discussion
id: create-discussion
continue-on-error: true
uses: actions/github-script@f28e40c7f34bde8b3046d885e986cb6290c5673b # v7.1.0
with:
script: |
const fs = require('fs');
const title = '${{ steps.report.outputs.title }}'.replace(/'/g, "\\'");
const body = fs.readFileSync('report-body.md', 'utf8');
// Find the "Skill Quality Reports" category
const categoriesResult = await github.graphql(`
query($owner: String!, $repo: String!) {
repository(owner: $owner, name: $repo) {
id
discussionCategories(first: 25) {
nodes { id name }
}
}
}
`, {
owner: context.repo.owner,
repo: context.repo.repo,
});
const repo = categoriesResult.repository;
const categories = repo.discussionCategories.nodes;
const category = categories.find(c =>
c.name === 'Skill Quality Reports'
);
if (!category) {
core.setFailed('Discussion category "Skill Quality Reports" not found. Falling back to issue.');
return;
}
await github.graphql(`
mutation($repoId: ID!, $categoryId: ID!, $title: String!, $body: String!) {
createDiscussion(input: {
repositoryId: $repoId,
categoryId: $categoryId,
title: $title,
body: $body
}) {
discussion { url }
}
}
`, {
repoId: repo.id,
categoryId: category.id,
title: title,
body: body,
});
console.log('Discussion created successfully.');
- name: Fallback — Create Issue
if: steps.create-discussion.outcome == 'failure'
env:
GH_TOKEN: ${{ github.token }}
run: |
# Create label if it doesn't exist (ignore errors if it already exists)
gh label create "skill-quality" --description "Automated skill quality reports" --color "d4c5f9" 2>/dev/null || true
gh issue create \
--title "${{ steps.report.outputs.title }}" \
--body-file report-body.md \
--label "skill-quality"