[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6J4C6ZPZYXGDVAWNHHXY2SD":3,"skill-same:01M6J4C6ZPZYXGDVAWNHHXY2SD":79},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":22,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":48,"commitSha":49,"skillDirectory":50,"forksCount":51,"repositoryHtmlUrl":52,"repositoryHomepageUrl":6,"ownerHtmlUrl":53,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":54,"install":55,"resources":68},"01M6J4C6ZPZYXGDVAWNHHXY2SD","mcs-eval",null,"Use this skill to test an agent's quality after building or fixing it. Runs eval test sets against a published agent via Direct Line API (default, auto), CopilotStudio SDK, Power CAT Kit, or MCS Native Eval (for MCP agents). Results are written per-test to evalSets[].tests[].lastResult in agentspec.json. Use after \u002Fmcs-build or \u002Fmcs-fix, or when the user wants to re-test.","dayour\u002FMCS-Agent-Builder","dayour","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F27440189?v=4","01M6J4B4HCG47B9AN14ADC3FV5","01M6J4B4P4QYVAZ0C7PZ84E1XD",0,"MIT","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fmcs-eval\u002FSKILL.md",[17],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"mcs eval","eval","power","for","kit","re","native","line","cat","tests","default","MCP","auto","use","fix","results","direct","api","copilotstudio","test","2026-11-10 10:12:30",false,"---\nname: mcs-eval\ndescription: \"Use this skill to test an agent's quality after building or fixing it. Runs eval test sets against a published agent via Direct Line API (default, auto), CopilotStudio SDK, Power CAT Kit, or MCS Native Eval (for MCP agents). Results are written per-test to evalSets[].tests[].lastResult in agentspec.json. Use after \u002Fmcs-build or \u002Fmcs-fix, or when the user wants to re-test.\"\n---\n\n# MCS Evaluation Runner\n\nRun evaluation tests for an agent and write results back to `agentspec.json` so the dashboard can display them.\n\n## Eval Runners — Default is Direct Line\n\n| Runner | When | Speed |\n|--------|------|-------|\n| **Direct Line** (default) | Agent has no user-delegated MCP tools | Fast (~2s\u002Ftest) |\n| **Direct Line + Interactive** | Agent requires OAuth sign-in | ~30s\u002Ftest + user input |\n| **CopilotStudio SDK** | User specifies `--sdk` | ~3s\u002Ftest |\n| **Power CAT Kit** | User specifies `--powercat` | Slow (server-side) |\n| **MCS Native Eval** | Agent uses MCP\u002Fuser-delegated tools, or `--native` | User-driven |\n\n**Auto-detection:** MCP\u002Fuser-delegated tools → MCS Native Eval. Otherwise → Direct Line. See `reference\u002Falternative-runners.md` for SDK, Power CAT Kit, and Native Eval details.\n\n## Input\n\n```\n\u002Fmcs-eval {projectId} {agentId}                          # Run all eval sets\n\u002Fmcs-eval {projectId} {agentId} --set boundaries,quality  # Run specific sets\n\u002Fmcs-eval {projectId} {agentId} --native                  # Force MCS Native Eval\n\u002Fmcs-eval {projectId} {agentId} --sdk                     # Force CopilotStudio SDK\n\u002Fmcs-eval {projectId} {agentId} --powercat                # Force Power CAT Kit\n\u002Fmcs-eval {projectId} {agentId} --check-results           # Check pending native eval results\n\u002Fmcs-eval {projectId} {agentId} --background              # Run the eval runner via Agent(run_in_background=true) so the lead session continues working; results land in agentspec.json on completion\n```\n\n**When to use `--background`:** large eval sets (40+ tests), MCS Native Eval (user-driven, can take 10-30 min), or any time you have other work to do while tests run. The Agent tool returns immediately; the lead is notified when the runner completes. Skip background mode for short Direct Line runs (\u003C3 min) — the overhead isn't worth it.\n\nReads from: `Build-Guides\u002F{projectId}\u002Fagents\u002F{agentId}\u002Fagentspec.json` — evalSets + buildStatus\nWrites to: `agentspec.json` (evalSets[].tests[].lastResult), `evals-results.json`, `evals-{setName}.csv`\n\n## Prerequisites: Auth Verification\n\nRe-verify auth from `\u002Fmcs-build`. Confirm account + environment with user (one-line). Azure CLI must match `buildStatus.azTenantId`, Dataverse must be reachable. If missing → \"Run `\u002Fmcs-build` first.\"\n\n## Before Evaluating — Knowledge + Learnings\n\n1. Read `knowledge\u002Fcache\u002Feval-methods.md` — refresh if stale (> 3 days)\n2. Read `knowledge\u002Flearnings\u002Feval-testing.md` for known failure patterns, scoring calibration, test design lessons\n\n## Step 0.5: Pre-Flight Checks\n\nVerify the agent is ready to be evaluated. These catch the #1 mistake: running eval on an unfinished agent.\n\n1. **Agent is published** — `buildStatus.status` must be `\"published\"`. If not → **STOP.**\n2. **Instructions exist** — must be non-empty. Under 50 chars → **WARN.**\n3. **Tools connected** (if configured) — match `integrations[]` against `buildStatus.toolsConfigured[]`. Missing → **WARN.**\n4. **Knowledge sources indexed** (if configured) — check `knowledge[]` against `buildStatus.knowledgeConfigured[]`. Missing → **WARN.**\n\nAny STOP → exit. Any WARN → list warnings, ask user to continue.\n\n## Step 1: Load Eval Sets & Determine Scope\n\nRead `agentspec.json.evalSets[]`. If empty → **exit:** \"Run `\u002Fmcs-research` first.\"\n\nDefault: run ALL sets. `--set boundaries,quality`: run only named sets. Skip sets with zero tests.\n\nGenerate per-set CSVs (`evals-boundaries.csv`, etc.) for dashboard download\u002Freference. Max 100 questions per CSV.\n\n## Step 1.5: Auto-Mode Detection\n\nCheck `agentspec.json.integrations[]` for MCP\u002Fuser-delegated tools:\n- **MCP\u002Fuser-delegated found** → MCS Native Eval (see `reference\u002Falternative-runners.md`)\n- **No user-delegated tools** → Direct Line (proceed to Step 2)\n\nUser can override with `--native`, `--sdk`, or `--powercat`.\n\n## Step 2: Acquire Direct Line Token\n\nTry in order (stop at first success):\n1. **Cached token endpoint** — `buildStatus.tokenEndpoint` if already discovered (tokens are acquired fresh each run — never persist bearer tokens in agentspec.json)\n2. **Token Endpoint discovery** — GET token endpoint URL → returns `{ Token, Expires_in, ConversationId }`. Cache the endpoint URL (not the token) in buildStatus.\n3. **Dataverse bound action** — `PvaGetDirectLineEndpoint` on bot entity\n4. **Ask user** — Copy from MCS UI: Settings → Security → Web channel security\n5. **All failed** → fall back to MCS Native Eval\n\n## Step 3: Run Tests — Direct Line\n\n```bash\n# With Token Endpoint (preferred — auto-refreshes)\nnode tools\u002Fdirect-line-test.js --token-endpoint \"\u003CURL>\" --csv \"Build-Guides\u002F{projectId}\u002Fagents\u002F{agentId}\u002Fevals-{setName}.csv\" --verbose\n\n# With manual token\nnode tools\u002Fdirect-line-test.js --token \"\u003CTOKEN>\" --csv \"Build-Guides\u002F{projectId}\u002Fagents\u002F{agentId}\u002Fevals-{setName}.csv\" --verbose\n\n# Custom timeout for slow agents\nnode tools\u002Fdirect-line-test.js --token-endpoint \"\u003CURL>\" --csv \"evals-{setName}.csv\" --timeout 90000 --verbose\n```\n\n**`--check-results` flow (native eval only):** When native eval was used previously, `--check-results` fetches pending results from the MCS eval dashboard via Gateway API `makerevaluations\u002Ftestcomponent\u002F{id}`, maps them back to `evalSets[].tests[].lastResult`, and writes to agentspec.json. Does not re-run tests.\n\n**Partial results:** If exit code 2 and > 50% completed → report partial results. If \u003C 50% → fall back to MCS Native Eval.\n\nResults saved to `evals-results.json` with `status`, `summary`, `method`, `results[]`.\n\n## Step 4: Write Results to agentspec.json\n\nUpdate `evalSets[].tests[].lastResult` for each test:\n```json\n{ \"pass\": true, \"actual\": \"...\", \"score\": 85, \"timestamp\": \"2026-02-18T14:30:00Z\" }\n```\n\nDo NOT write a flat `evalResults` field — results live per-test in their eval set. Cache token endpoint URL (not the token itself) in buildStatus.\n\n**VERIFY:** Read agentspec.json back. Confirm each test has `lastResult` with `pass`, `actual`, and `timestamp`.\n\n**Borderline tests:** For tests within 15 points of threshold, GPT dual scoring activates automatically. See `reference\u002Fdual-scoring.md` for the full protocol.\n\n## Step 5: Report Results + Verdict\n\nAfter scoring, compute the SHIP\u002FITERATE\u002FBLOCK verdict using the eval-guide risk-based model (stored in `evalConfig.lastVerdict` by the eval pipeline).\n\nThen invoke the eval-guide plugin `\u002Feval-result-interpreter` for full interpretation:\n- Pass the per-test results (or exported CSV) to the plugin\n- The plugin produces: structured triage report, root cause classification, top 3 remediation actions, pattern analysis\n- Present the verdict prominently, followed by the per-set breakdown\n\n```\n## Evaluation Results: {Agent Name}\n\n**Verdict:** {SHIP | SHIP WITH KNOWN GAPS | ITERATE | BLOCK}\n**Method:** {Direct Line API | MCS Native Eval | Manual Test Chat}\n**Sets run:** {set names}\n**Overall:** {X}\u002F{Y} passed ({Z}%)\n\n**Per-Set Results (eval-guide risk-based thresholds):**\n| Set | Passed | Total | Rate | Target | Block At | Status |\n| boundaries | X | Y | Z% | >=95% | \u003C95% | PASS\u002FBLOCK |\n| quality | X | Y | Z% | >=90% | \u003C80% | PASS\u002FITERATE\u002FBLOCK |\n| edge-cases | X | Y | Z% | >=70% | — | PASS\u002FITERATE |\n\n**Failed Cases:**\n| Set | Question | Expected | Got | Root Cause Type |\n\n**Top 3 Actions (from eval-guide interpreter):**\n1. Change X → Re-run Y → Expect Z\n2. ...\n3. ...\n```\n\nIf the eval-guide plugin is unavailable, fall back to the programmatic verdict in `evalConfig.lastVerdict` (computed by eval-pipeline.js).\n\n## Quality Standards (eval-guide risk-based model)\n\n- **Safety\u002Fcompliance (boundaries) >=95%** — below = BLOCK, fix before anything else\n- **Core business (quality) >=90% target, \u003C80% = BLOCK** — below target = ITERATE\n- **Edge cases >=70% target** — iterate, don't block (hard by design)\n- **Overall \u003C60% = BLOCK** regardless of per-set results\n- **Re-run eval after any agent changes**\n- **GeneralQuality evals have variance** — run multiple times for confidence (+\u002F-5% is normal)\n- **100% pass rate is a red flag** — eval is likely too easy, add harder cases\n\n## Post-Eval Learnings\n\nTier 1 (auto): bump `confirmed` counts for matching learnings entries. Tier 2 (user-confirmed): capture new failure patterns, method insights, scoring calibration. Run comparison engine before writing new entries. See `.claude\u002Frules\u002Flearnings-system.md` for protocol.\n\n## Gotchas\n\n- **agentspec.json evalSets is the primary output** — the dashboard reads per-test lastResult from it\n- **Never mark eval complete after only uploading** — run the evaluation and write per-test results\n- **Manual mode uploads via Gateway API** — CSVs are for dashboard download only, not for import\n- **Cache the token endpoint URL** in agentspec.json for future eval runs\n- **Per-set pass logic:** each test must pass ALL methods in its set\n- **Token Endpoint is the most reliable method** — no manual steps needed\n","# MCS Evaluation Runner\n\nRun evaluation tests for an agent and write results back to `agentspec.json` so the dashboard can display them.\n\n## Eval Runners — Default is Direct Line\n\n| Runner | When | Speed |\n|--------|------|-------|\n| **Direct Line** (default) | Agent has no user-delegated MCP tools | Fast (~2s\u002Ftest) |\n| **Direct Line + Interactive** | Agent requires OAuth sign-in | ~30s\u002Ftest + user input |\n| **CopilotStudio SDK** | User specifies `--sdk` | ~3s\u002Ftest |\n| **Power CAT Kit** | User specifies `--powercat` | Slow (server-side) |\n| **MCS Native Eval** | Agent uses MCP\u002Fuser-delegated tools, or `--native` | User-driven |\n\n**Auto-detection:** MCP\u002Fuser-delegated tools → MCS Native Eval. Otherwise → Direct Line. See `reference\u002Falternative-runners.md` for SDK, Power CAT Kit, and Native Eval details.\n\n## Input\n\n```\n\u002Fmcs-eval {projectId} {agentId}                          # Run all eval sets\n\u002Fmcs-eval {projectId} {agentId} --set boundaries,quality  # Run specific sets\n\u002Fmcs-eval {projectId} {agentId} --native                  # Force MCS Native Eval\n\u002Fmcs-eval {projectId} {agentId} --sdk                     # Force CopilotStudio SDK\n\u002Fmcs-eval {projectId} {agentId} --powercat                # Force Power CAT Kit\n\u002Fmcs-eval {projectId} {agentId} --check-results           # Check pending native eval results\n\u002Fmcs-eval {projectId} {agentId} --background              # Run the eval runner via Agent(run_in_background=true) so the lead session continues working; results land in agentspec.json on completion\n```\n\n**When to use `--background`:** large eval sets (40+ tests), MCS Native Eval (user-driven, can take 10-30 min), or any time you have other work to do while tests run. The Agent tool returns immediately; the lead is notified when the runner completes. Skip background mode for short Direct Line runs (\u003C3 min) — the overhead isn't worth it.\n\nReads from: `Build-Guides\u002F{projectId}\u002Fagents\u002F{agentId}\u002Fagentspec.json` — evalSets + buildStatus\nWrites to: `agentspec.json` (evalSets[].tests[].lastResult), `evals-results.json`, `evals-{setName}.csv`\n\n## Prerequisites: Auth Verification\n\nRe-verify auth from `\u002Fmcs-build`. Confirm account + environment with user (one-line). Azure CLI must match `buildStatus.azTenantId`, Dataverse must be reachable. If missing → \"Run `\u002Fmcs-build` first.\"\n\n## Before Evaluating — Knowledge + Learnings\n\n1. Read `knowledge\u002Fcache\u002Feval-methods.md` — refresh if stale (> 3 days)\n2. Read `knowledge\u002Flearnings\u002Feval-testing.md` for known failure patterns, scoring calibration, test design lessons\n\n## Step 0.5: Pre-Flight Checks\n\nVerify the agent is ready to be evaluated. These catch the #1 mistake: running eval on an unfinished agent.\n\n1. **Agent is published** — `buildStatus.status` must be `\"published\"`. If not → **STOP.**\n2. **Instructions exist** — must be non-empty. Under 50 chars → **WARN.**\n3. **Tools connected** (if configured) — match `integrations[]` against `buildStatus.toolsConfigured[]`. Missing → **WARN.**\n4. **Knowledge sources indexed** (if configured) — check `knowledge[]` against `buildStatus.knowledgeConfigured[]`. Missing → **WARN.**\n\nAny STOP → exit. Any WARN → list warnings, ask user to continue.\n\n## Step 1: Load Eval Sets & Determine Scope\n\nRead `agentspec.json.evalSets[]`. If empty → **exit:** \"Run `\u002Fmcs-research` first.\"\n\nDefault: run ALL sets. `--set boundaries,quality`: run only named sets. Skip sets with zero tests.\n\nGenerate per-set CSVs (`evals-boundaries.csv`, etc.) for dashboard download\u002Freference. Max 100 questions per CSV.\n\n## Step 1.5: Auto-Mode Detection\n\nCheck `agentspec.json.integrations[]` for MCP\u002Fuser-delegated tools:\n- **MCP\u002Fuser-delegated found** → MCS Native Eval (see `reference\u002Falternative-runners.md`)\n- **No user-delegated tools** → Direct Line (proceed to Step 2)\n\nUser can override with `--native`, `--sdk`, or `--powercat`.\n\n## Step 2: Acquire Direct Line Token\n\nTry in order (stop at first success):\n1. **Cached token endpoint** — `buildStatus.tokenEndpoint` if already discovered (tokens are acquired fresh each run — never persist bearer tokens in agentspec.json)\n2. **Token Endpoint discovery** — GET token endpoint URL → returns `{ Token, Expires_in, ConversationId }`. Cache the endpoint URL (not the token) in buildStatus.\n3. **Dataverse bound action** — `PvaGetDirectLineEndpoint` on bot entity\n4. **Ask user** — Copy from MCS UI: Settings → Security → Web channel security\n5. **All failed** → fall back to MCS Native Eval\n\n## Step 3: Run Tests — Direct Line\n\n```bash\n# With Token Endpoint (preferred — auto-refreshes)\nnode tools\u002Fdirect-line-test.js --token-endpoint \"\u003CURL>\" --csv \"Build-Guides\u002F{projectId}\u002Fagents\u002F{agentId}\u002Fevals-{setName}.csv\" --verbose\n\n# With manual token\nnode tools\u002Fdirect-line-test.js --token \"\u003CTOKEN>\" --csv \"Build-Guides\u002F{projectId}\u002Fagents\u002F{agentId}\u002Fevals-{setName}.csv\" --verbose\n\n# Custom timeout for slow agents\nnode tools\u002Fdirect-line-test.js --token-endpoint \"\u003CURL>\" --csv \"evals-{setName}.csv\" --timeout 90000 --verbose\n```\n\n**`--check-results` flow (native eval only):** When native eval was used previously, `--check-results` fetches pending results from the MCS eval dashboard via Gateway API `makerevaluations\u002Ftestcomponent\u002F{id}`, maps them back to `evalSets[].tests[].lastResult`, and writes to agentspec.json. Does not re-run tests.\n\n**Partial results:** If exit code 2 and > 50% completed → report partial results. If \u003C 50% → fall back to MCS Native Eval.\n\nResults saved to `evals-results.json` with `status`, `summary`, `method`, `results[]`.\n\n## Step 4: Write Results to agentspec.json\n\nUpdate `evalSets[].tests[].lastResult` for each test:\n```json\n{ \"pass\": true, \"actual\": \"...\", \"score\": 85, \"timestamp\": \"2026-02-18T14:30:00Z\" }\n```\n\nDo NOT write a flat `evalResults` field — results live per-test in their eval set. Cache token endpoint URL (not the token itself) in buildStatus.\n\n**VERIFY:** Read agentspec.json back. Confirm each test has `lastResult` with `pass`, `actual`, and `timestamp`.\n\n**Borderline tests:** For tests within 15 points of threshold, GPT dual scoring activates automatically. See `reference\u002Fdual-scoring.md` for the full protocol.\n\n## Step 5: Report Results + Verdict\n\nAfter scoring, compute the SHIP\u002FITERATE\u002FBLOCK verdict using the eval-guide risk-based model (stored in `evalConfig.lastVerdict` by the eval pipeline).\n\nThen invoke the eval-guide plugin `\u002Feval-result-interpreter` for full interpretation:\n- Pass the per-test results (or exported CSV) to the plugin\n- The plugin produces: structured triage report, root cause classification, top 3 remediation actions, pattern analysis\n- Present the verdict prominently, followed by the per-set breakdown\n\n```\n## Evaluation Results: {Agent Name}\n\n**Verdict:** {SHIP | SHIP WITH KNOWN GAPS | ITERATE | BLOCK}\n**Method:** {Direct Line API | MCS Native Eval | Manual Test Chat}\n**Sets run:** {set names}\n**Overall:** {X}\u002F{Y} passed ({Z}%)\n\n**Per-Set Results (eval-guide risk-based thresholds):**\n| Set | Passed | Total | Rate | Target | Block At | Status |\n| boundaries | X | Y | Z% | >=95% | \u003C95% | PASS\u002FBLOCK |\n| quality | X | Y | Z% | >=90% | \u003C80% | PASS\u002FITERATE\u002FBLOCK |\n| edge-cases | X | Y | Z% | >=70% | — | PASS\u002FITERATE |\n\n**Failed Cases:**\n| Set | Question | Expected | Got | Root Cause Type |\n\n**Top 3 Actions (from eval-guide interpreter):**\n1. Change X → Re-run Y → Expect Z\n2. ...\n3. ...\n```\n\nIf the eval-guide plugin is unavailable, fall back to the programmatic verdict in `evalConfig.lastVerdict` (computed by eval-pipeline.js).\n\n## Quality Standards (eval-guide risk-based model)\n\n- **Safety\u002Fcompliance (boundaries) >=95%** — below = BLOCK, fix before anything else\n- **Core business (quality) >=90% target, \u003C80% = BLOCK** — below target = ITERATE\n- **Edge cases >=70% target** — iterate, don't block (hard by design)\n- **Overall \u003C60% = BLOCK** regardless of per-set results\n- **Re-run eval after any agent changes**\n- **GeneralQuality evals have variance** — run multiple times for confidence (+\u002F-5% is normal)\n- **100% pass rate is a red flag** — eval is likely too easy, add harder cases\n\n## Post-Eval Learnings\n\nTier 1 (auto): bump `confirmed` counts for matching learnings entries. Tier 2 (user-confirmed): capture new failure patterns, method insights, scoring calibration. Run comparison engine before writing new entries. See `.claude\u002Frules\u002Flearnings-system.md` for protocol.\n\n## Gotchas\n\n- **agentspec.json evalSets is the primary output** — the dashboard reads per-test lastResult from it\n- **Never mark eval complete after only uploading** — run the evaluation and write per-test results\n- **Manual mode uploads via Gateway API** — CSVs are for dashboard download only, not for import\n- **Cache the token endpoint URL** in agentspec.json for future eval runs\n- **Per-set pass logic:** each test must pass ALL methods in its set\n- **Token Endpoint is the most reliable method** — no manual steps needed\n",{"name":5,"description":7},"main","b7593fb57652591678d1aeb9bcc19738e903ffbe",".claude\u002Fskills\u002Fmcs-eval",3,"https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder","https:\u002F\u002Fgithub.com\u002Fdayour","2026-11-08 11:20:11",{"source":8,"skillName":5,"commands":56,"prompt":66,"zipAvailable":67},[57,60,63],{"tool":58,"command":59},"npx","npx skills add dayour\u002FMCS-Agent-Builder --skill mcs-eval",{"tool":61,"command":62},"bunx","bunx skills add dayour\u002FMCS-Agent-Builder --skill mcs-eval",{"tool":64,"command":65},"pnpm","pnpm dlx skills add dayour\u002FMCS-Agent-Builder --skill mcs-eval","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fmcs-eval\u002FSKILL.md",true,[69,75],{"path":70,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":73,"sourceHtmlUrl":74},"reference\u002Falternative-runners.md","REFERENCE","参考文档",3807,"https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fmcs-eval\u002Freference\u002Falternative-runners.md",{"path":76,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":77,"sourceHtmlUrl":78},"reference\u002Fdual-scoring.md",1540,"https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fmcs-eval\u002Freference\u002Fdual-scoring.md",[80,99,106,111,118,123],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":43},"01M6WTHH2JPKXM1JDRE55KC4A6","order-status","Read the packaged Copilot Studio agent definition, including messaging topics and boundaries, from the local plugin.","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002Fschema\u002Ffixtures\u002Finvalid\u002Fmodel-secret\u002Fskills\u002Forder-status\u002FSKILL.md",[],[87,88,89,90,91,92,93,94,95,96,97,98],"order status","status","copilot","plugin","copilot studio","messaging","read","local","agent definition","studio","definition","topics",{"skillNo":100,"name":101,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":43},"01M6WTHE20Y6X94754CP3TAY1Y","level-pass","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002Fschema\u002Ffixtures\u002Finvalid\u002Flevel-status-failed\u002Fskills\u002Flevel-pass\u002FSKILL.md",[],[105,93,92,91,90,89,96,94,95,97,98],"level pass",{"skillNo":107,"name":101,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":110,"publishedAt":43},"01M6WTHBZ5EY53QPYKREMEVHZK","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002Fschema\u002Ffixtures\u002Finvalid\u002Flevel-status-fail\u002Fskills\u002Flevel-pass\u002FSKILL.md",[],[105,97,98,96,95,94,93,92,89,91,90],{"skillNo":112,"name":113,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":114,"categories":115,"keywords":116,"publishedAt":43},"01M6WTH9AKX173Y69CJS3CZM5A","level-not-passing","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002Fschema\u002Ffixtures\u002Finvalid\u002Flevel-malformed\u002Fskills\u002Flevel-not-passing\u002FSKILL.md",[],[117,89,90,91,92,93,94,95,96,97,98],"level not passing",{"skillNo":119,"name":101,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":120,"categories":121,"keywords":122,"publishedAt":43},"01M6WTH70Z5Z2JV26J9C1ZRYPS","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002Fschema\u002Ffixtures\u002Finvalid\u002Flevel-long-fence\u002Fskills\u002Flevel-pass\u002FSKILL.md",[],[105,95,89,90,91,92,93,94,96,97,98],{"skillNo":124,"name":101,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":9,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":125,"categories":126,"keywords":127,"publishedAt":43},"01M6WTH4ZSAJ8SADD803KY2DW0","https:\u002F\u002Fgithub.com\u002Fdayour\u002FMCS-Agent-Builder\u002Fblob\u002Fmain\u002Fschema\u002Ffixtures\u002Finvalid\u002Flevel-indented\u002Fskills\u002Flevel-pass\u002FSKILL.md",[],[105,90,91,92,93,94,95,96,97,98,89]]