mirror of
https://github.com/proffesor-for-testing/agentic-qe.git
synced 2026-09-19 08:45:47 +08:00
Dragan Spiridonov
25b6bd732a
Bundles 4 MCP bug fixes + new agentic-qe-fleet plugin. Fixed: - coverage_analyze_sublinear and other domain MCP tools blocked by governance after 1-2 calls (synthetic token-slope estimate replaced with 0; throttle decisions are no longer treated as hard blocks) - test_generate_enhanced framework=jest produced unrunnable tests (now emits @jest/globals imports instead of invalid 'jest' package import) - generated test imports referenced /tmp/aqe-temp-* paths (importPathOverrides plumbed through IGenerateTestsRequest end-to-end) - qe/coverage/gaps error message misled callers about which input failed (distinguishes coverageFile-empty vs autodiscovery-miss) Added: - test_generate_enhanced schema exposes framework, filePath, coverageGoal, aiEnhancement, detectAntiPatterns parameters - agentic-qe-fleet Claude Code plugin (11 agents, 9 commands, 9 skills) - scripts/smoke-test-fixes.sh and scripts/plugin-load-test.sh Verification: - 1301 MCP unit tests pass (zero regressions) - 734 targeted unit tests pass across changed code - aqe init --auto in fresh project succeeds (85 skills, 60 agents) - End-to-end MCP smoke test passes 5/5 - qe-coverage-specialist agent re-spawn confirms NO_GOVERNANCE_BLOCK - Plugin structural validation: 14/14 checks pass Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
111 lines
3.8 KiB
Bash
Executable File
111 lines
3.8 KiB
Bash
Executable File
#!/bin/bash
|
|
# Plugin load test — validates that the bundled plugin actually works
|
|
# in a fresh Claude Code session. Must be run from a terminal (not from
|
|
# within an existing Claude Code session).
|
|
#
|
|
# This script does the structural checks that don't require running Claude.
|
|
# For the full end-to-end check, run:
|
|
# claude --plugin-dir ./plugins/agentic-qe-fleet
|
|
# and verify that:
|
|
# - The 9 /aqe-* slash commands appear in the list
|
|
# - The 11 qe-* agents are spawnable
|
|
# - The 9 skills are listed
|
|
# - The agentic-qe MCP server connects (claude mcp list shows it)
|
|
|
|
set -u
|
|
PLUGIN=plugins/agentic-qe-fleet
|
|
|
|
if [ ! -d "$PLUGIN" ]; then
|
|
echo "FAIL plugin directory not found: $PLUGIN" >&2
|
|
exit 1
|
|
fi
|
|
|
|
passes=0
|
|
fails=0
|
|
|
|
check() {
|
|
local label="$1" cond="$2"
|
|
if eval "$cond"; then
|
|
echo "PASS $label"
|
|
passes=$((passes + 1))
|
|
else
|
|
echo "FAIL $label"
|
|
fails=$((fails + 1))
|
|
fi
|
|
}
|
|
|
|
# === Structural checks ===
|
|
check "plugin.json exists" "[ -f $PLUGIN/.claude-plugin/plugin.json ]"
|
|
check "plugin.json valid JSON" "node -e 'require(\"./$PLUGIN/.claude-plugin/plugin.json\")' 2>/dev/null"
|
|
check ".mcp.json exists" "[ -f $PLUGIN/.mcp.json ]"
|
|
check ".mcp.json valid JSON" "node -e 'require(\"./$PLUGIN/.mcp.json\")' 2>/dev/null"
|
|
check "README.md present" "[ -f $PLUGIN/README.md ]"
|
|
|
|
# === Counts ===
|
|
agents=$(ls $PLUGIN/agents/*.md 2>/dev/null | wc -l)
|
|
commands=$(ls $PLUGIN/commands/*.md 2>/dev/null | wc -l)
|
|
skills=$(ls -d $PLUGIN/skills/*/ 2>/dev/null | wc -l)
|
|
check "agents count = 11" "[ $agents -eq 11 ]"
|
|
check "commands count = 9" "[ $commands -eq 9 ]"
|
|
check "skills count = 9" "[ $skills -eq 9 ]"
|
|
|
|
# === Frontmatter ===
|
|
fm_errors=$(node -e "
|
|
const fs = require('fs'), path = require('path');
|
|
let errors = 0;
|
|
function walk(d) {
|
|
for (const e of fs.readdirSync(d, { withFileTypes: true })) {
|
|
const p = path.join(d, e.name);
|
|
if (e.isDirectory()) walk(p);
|
|
else if (e.name.endsWith('.md')) {
|
|
const c = fs.readFileSync(p, 'utf8');
|
|
if (!c.startsWith('---')) continue;
|
|
const i = c.indexOf('---', 3);
|
|
if (i < 0) { errors++; continue; }
|
|
if (!/^name:/m.test(c.slice(3, i))) errors++;
|
|
}
|
|
}
|
|
}
|
|
walk('$PLUGIN');
|
|
console.log(errors);
|
|
" 2>/dev/null)
|
|
check "all .md files have valid frontmatter" "[ \"$fm_errors\" = '0' ]"
|
|
|
|
# === Each agent has a model field ===
|
|
agents_no_model=$(grep -L "^model:" $PLUGIN/agents/*.md 2>/dev/null | wc -l)
|
|
check "all agents have model field" "[ $agents_no_model -eq 0 ]"
|
|
|
|
# === Each skill has trust_tier and allowed-tools ===
|
|
skills_no_tier=0
|
|
skills_no_tools=0
|
|
for s in $PLUGIN/skills/*/SKILL.md; do
|
|
if ! grep -q "^trust_tier:" "$s"; then skills_no_tier=$((skills_no_tier + 1)); fi
|
|
if ! grep -q "^allowed-tools:" "$s"; then skills_no_tools=$((skills_no_tools + 1)); fi
|
|
done
|
|
check "all skills have trust_tier" "[ $skills_no_tier -eq 0 ]"
|
|
check "all skills have allowed-tools" "[ $skills_no_tools -eq 0 ]"
|
|
|
|
# === MCP launcher resolves ===
|
|
if command -v aqe-mcp >/dev/null 2>&1; then
|
|
check "aqe-mcp on PATH (published bin)" "true"
|
|
elif command -v agentic-qe >/dev/null 2>&1; then
|
|
check "agentic-qe on PATH (published bin)" "true"
|
|
else
|
|
check "MCP launcher on PATH" "false"
|
|
fi
|
|
|
|
# === No tier-1 skills shipped ===
|
|
tier1_skills=$(grep -l "^trust_tier: 1" $PLUGIN/skills/*/SKILL.md 2>/dev/null | wc -l)
|
|
check "no tier-1 skills shipped" "[ $tier1_skills -eq 0 ]"
|
|
|
|
echo ""
|
|
echo "=== $passes passed, $fails failed ==="
|
|
echo ""
|
|
echo "Manual end-to-end load test (must be run from a fresh terminal):"
|
|
echo " claude --plugin-dir ./plugins/agentic-qe-fleet"
|
|
echo "Then in the loaded session:"
|
|
echo " /aqe-fleet-status # commands resolve"
|
|
echo " claude mcp list # agentic-qe shows as Connected"
|
|
echo " Task tool: spawn qe-test-architect agent"
|
|
exit $((fails == 0 ? 0 : 1))
|