feat(lessons): 1 community lessons (heartbeat batch)

data/leaderboard.json

@@ -1,134 +1,150 @@
 [
   {
     "login": "zsxh1990",
-    "score": 22.75
+    "score": 24.6
+  },
+  {
+    "login": "uncledad96-glitch",
+    "score": 15.26
   },
   {
     "login": "2lll5",
-    "score": 6.15
+    "score": 5.02
   },
   {
     "login": "solaris",
-    "score": 3.44
+    "score": 3.05
+  },
+  {
+    "login": "lb1192176991-lab",
+    "score": 1.87
   },
   {
     "login": "zeroknowledge0x",
-    "score": 1.82
+    "score": 1.55
   },
   {
     "login": "sparshgarg999",
-    "score": 1.42
+    "score": 1.26
+  },
+  {
+    "login": "namdamdoi68-oss",
+    "score": 0.97
+  },
+  {
+    "login": "ninghuagui-debug",
+    "score": 0.94
   },
   {
     "login": "yh-liao-07",
-    "score": 1.0
+    "score": 0.93
   },
   {
     "login": "abhiavi",
-    "score": 1.0
+    "score": 0.93
   },
   {
     "login": "ringotokens-commits",
-    "score": 0.99
+    "score": 0.93
   },
   {
     "login": "ivegotahunnitonit",
-    "score": 0.99
+    "score": 0.93
   },
   {
     "login": "root",
-    "score": 0.99
+    "score": 0.62
   },
   {
     "login": "lushan888",
-    "score": 0.95
+    "score": 0.58
   },
   {
     "login": "chiranjeevi7777",
-    "score": 0.95
+    "score": 0.58
   },
   {
     "login": "zhangjianming7051",
-    "score": 0.95
+    "score": 0.58
   },
   {
     "login": "xkes101116-hub",
-    "score": 0.94
+    "score": 0.58
   },
   {
     "login": "foxymantou",
-    "score": 0.94
+    "score": 0.58
   },
   {
     "login": "votienduong2208",
-    "score": 0.65
+    "score": 0.56
   },
   {
     "login": "syzygys",
-    "score": 0.59
+    "score": 0.53
   },
   {
     "login": "jhon12091986",
-    "score": 0.59
+    "score": 0.53
   },
   {
     "login": "lovewave02",
-    "score": 0.58
+    "score": 0.52
   },
   {
     "login": "gothundercats",
-    "score": 0.56
+    "score": 0.51
   },
   {
     "login": "zqleslie",
-    "score": 0.55
+    "score": 0.49
   },
   {
     "login": "andrianbalanesq",
-    "score": 0.52
+    "score": 0.48
   },
   {
     "login": "cyprerask",
-    "score": 0.51
-  },
-  {
-    "login": "rohitmulani63-ops",
-    "score": 0.47
-  },
-  {
-    "login": "skyjames777",
-    "score": 0.47
+    "score": 0.46
   },
   {
     "login": "doview1",
-    "score": 0.33
+    "score": 0.28
   },
   {
     "login": "suresh chouksey",
-    "score": 0.33
+    "score": 0.28
   },
   {
     "login": "iccccccccccccc",
-    "score": 0.33
+    "score": 0.28
   },
   {
     "login": "sagarmaurya64-ai",
-    "score": 0.32
+    "score": 0.27
+  },
+  {
+    "login": "rohitmulani63-ops",
+    "score": 0.23
+  },
+  {
+    "login": "skyjames777",
+    "score": 0.23
   },
   {
     "login": "pian0",
-    "score": 0.2
+    "score": 0.17
   },
   {
     "login": "qi574",
-    "score": 0.16
+    "score": 0.14
   },
   {
     "login": "cuongwf1711",
-    "score": 0.16
+    "score": 0.14
   },
   {
     "login": "lqkhanh295",
-    "score": 0.13
+    "score": 0.11
   }
 ]

data/lessons.json

@@ -1945,6 +1945,28 @@
     "status": "active",
     "verified": false
   },
+  {
+    "id": "glama-introspection-gap",
+    "title": "Glama Introspection Gap — Build Success ≠ Tools Registered",
+    "domain": "devops",
+    "tags": [
+      "glama",
+      "mcp",
+      "introspection",
+      "tools",
+      "registry"
+    ],
+    "summary": "After successfully building an MCP server on Glama, the tools don't appear in the Glama API or dashboard. Build success ≠ tools registered — they are separate a…",
+    "preview": "## Problem\n\nAfter successfully building an MCP server on Glama, the tools don't appear in the Glama API or dashboard. Build success ≠ tools registered — they are separate async processes.\n\n## Root Cause\n\nGlama's pipeline has two distinct steps:\n1. **Build** — Docker image creation (synchronous, returns immediately)\n2. **Introspection** — Runs the MCP server and calls `tools/list` (async, may take minutes to hours)\n\nThe build step can succeed while introspection fails silently. Common causes:\n- MCP server starts but doesn't respond to `tools/list` request\n- Server crashes during introspection\n- glama.json format issues (tools not detected)\n- Network timeout during introspection\n\n## Detection\n\n```bash\n# Check if tools are registered\ncurl -s \"https://glama.ai/api/mcp/v1/servers/OWNER/REPO\" | python3 -c \"\nimport json, sys\ndata = json.load(sys.stdin)\nprint(f'Tools: {len(data.get(\\\"tools\\\", []))}')\n\"\n```\n\nIf `tools: 0` but build succeeded, introspection failed.\n\n## Fix Action\n\n1. **Wait** — introspection is async, may take hours\n2. **Sync Server** — trigger Glama to re-read the repo\n3. **Rebuild** — force fresh introspection\n4. **Check glama.json** — must be minimal format (`$schema` + `maintainers` only)\n5. **Check Dockerfile** — ensure MCP server starts and responds to `initialize` request\n\n## Key Insight\n\n**glama.json is NOT for tool definitions.** Glama discovers tools via MCP introspection (calling `tools/list`), not from glama.json. The glama.json should only contain:\n```json\n{\n  \"$schema\": \"https://glama.ai/mcp/schemas/server.json\",\n  \"maintainers\": [\"username\"]\n}\n```\n\nComplex tool definitions in glama.json are ignored by Glama's introspection system.",
+    "url": "lessons/contrib/glama-introspection-gap.md",
+    "created": "2026-07-22",
+    "updated": "",
+    "validity_period_days": 365,
+    "environment_version": "",
+    "confidence": 0.5,
+    "status": "published",
+    "verified": false
+  },
   {
     "id": "gpt-sovits-hubert-16khz",
     "title": "gpt sovits hubert 16khz",
@@ -2649,6 +2671,28 @@
     "status": "active",
     "verified": true
   },
+  {
+    "id": "merge-probability-calibration",
+    "title": "Merge Probability Calibration — Honest Estimation vs Overfitting",
+    "domain": "devops",
+    "tags": [
+      "prediction",
+      "calibration",
+      "merge-rate",
+      "a-b-test",
+      "coach"
+    ],
+    "summary": "PR coaches that predict merge probability often overfit to training data or give misleadingly precise estimates. A 70% accuracy claim may not generalize to new …",
+    "preview": "## Problem\n\nPR coaches that predict merge probability often overfit to training data or give misleadingly precise estimates. A 70% accuracy claim may not generalize to new repos or PR types.\n\n## Root Cause\n\n1. **Base rate trap** — Most repos have low external merge rates (20-30%). Predicting \"medium risk\" for everything gives 70%+ accuracy but zero discrimination.\n\n2. **Signal confusion** — Signals like \"needs_preflight\" or \"large_repo\" are risk markers, not success predictors. A PR can have many negative signals and still merge if the maintainer wants it.\n\n3. **Content blindness** — Current coaches analyze metadata (title, body, files_changed) but not actual diff content. Two PRs with identical metadata can have completely different merge outcomes.\n\n## A/B Test Results (445 cases)\n\n| Metric | Value |\n|--------|-------|\n| Merged PRs mean probability | 0.32 |\n| Closed PRs mean probability | 0.30 |\n| Gap | +0.02 |\n| Discrimination | YES (but limited) |\n\n**Conclusion:** Merge probability can't be more accurate than the repo's base merge rate without understanding PR content quality.\n\n## Fix Action\n\n1. **Use repo merge rate as base** — the most honest starting point\n2. **Only adjust for discriminating signals** — merge_conflict (×0.3), duplicate (×0.1), maintainer_internal (×0.05)\n3. **Don't double-count** — signals already affect tier, don't also affect probability\n4. **Be transparent** — tell users \"this repo has 20% merge rate, your PR is slightly better than average\"\n\n## Prevention\n\n- Always A/B test predictions against actual outcomes\n- Use LORO (Leave-One-Repo-Out) validation to detect overfitting\n- Report confidence intervals, not point estimates\n- Accept that some uncertainty is irreducible (depends on maintainer mood, timing, etc.)",
+    "url": "lessons/contrib/merge-probability-calibration.md",
+    "created": "2026-07-22",
+    "updated": "",
+    "validity_period_days": 365,
+    "environment_version": "",
+    "confidence": 0.5,
+    "status": "published",
+    "verified": false
+  },
   {
     "id": "misakanet-heal-engine-bootstrap-workflow",
     "title": "MisakaNet --heal Engine Bootstrap Workflow",
@@ -3348,6 +3392,28 @@
     "status": "published",
     "verified": true
   },
+  {
+    "id": "restricted-interactions-repos",
+    "title": "Restricted Interactions Repos — External PRs Not Accepted",
+    "domain": "devops",
+    "tags": [
+      "github",
+      "restricted",
+      "external-pr",
+      "contribution",
+      "gatekeeping"
+    ],
+    "summary": "Some large open-source repos have GitHub \"restricted interactions\" enabled, which prevents non-collaborators from creating PRs, commenting, or even viewing cert…",
+    "preview": "## Problem\n\nSome large open-source repos have GitHub \"restricted interactions\" enabled, which prevents non-collaborators from creating PRs, commenting, or even viewing certain features. Contributors waste time preparing PRs that can never be submitted.\n\n## Root Cause\n\nGitHub allows repo admins to restrict interactions to collaborators only. This is separate from the standard fork-and-PR workflow. When enabled:\n- Non-collaborators cannot create PRs\n- Non-collaborators cannot comment on issues/PRs\n- The error message is generic: \"Interactions on this repository have been restricted to collaborators only\"\n\n## Detection\n\nBefore preparing a PR, check:\n```bash\n# Check if you can create a PR\ngh pr create --repo org/repo --title \"test\" --body \"test\" 2>&1 | grep \"restricted\"\n\n# Check if you can comment\ngh issue comment 1 --repo org/repo --body \"test\" 2>&1 | grep \"restricted\"\n```\n\n## Known Repos with Restricted Interactions\n\n- `encode/httpx` — Python HTTP library, no external PRs accepted\n- Some Grafana Labs repos — require signed commits + restricted interactions\n\n## Fix Action\n\n1. Check repo interaction permissions before investing time\n2. If restricted, look for alternative contribution channels (discussions, docs)\n3. If no alternatives, move to a different repo\n\n## Prevention\n\nAlways test repo permissions before preparing a PR:\n```bash\n# Quick permission check\ngh api repos/org/repo --jq '.permissions'\n```\n\nIf `pull` is false or interactions are restricted, don't invest time in PR preparation.",
+    "url": "lessons/contrib/restricted-interactions-repos.md",
+    "created": "2026-07-22",
+    "updated": "",
+    "validity_period_days": 365,
+    "environment_version": "",
+    "confidence": 0.5,
+    "status": "published",
+    "verified": false
+  },
   {
     "id": "sag-lite-data-quality-cleaning",
     "title": "sag-lite-data-quality-cleaning",