Add harness configuration fingerprints

[ccf]

Summary

• Adds harness configuration fingerprints to maturity analytics, grouping sessions by agent, permission mode, tool set, context signal presence, and customization state.
• Exposes fingerprint metrics in the API response: session/engineer count, success rate, 10-step compound reliability, leverage score, tool/customization/context counts, and top signals.
• Adds a Harness Intelligence UI table for the fingerprint catalog and updates the roadmap item to shipped.

Validation

• ruff check .
• ruff format --check .
• bandit -r src/ -c pyproject.toml
• PYTHONPATH=/Users/ccf/git/primer/src:/Users/ccf/git/primer pytest --import-mode=importlib -v --tb=short - 962 passed
• ./node_modules/.bin/eslint src/pages/maturity.tsx src/components/maturity/harness-fingerprint-table.tsx src/components/maturity/__tests__/harness-fingerprint-table.test.tsx src/components/maturity/__tests__/maturity-summary.test.tsx src/types/api.ts
• ./node_modules/.bin/tsc --noEmit
• ./node_modules/.bin/vitest run --run src/components/maturity/__tests__/harness-fingerprint-table.test.tsx src/components/maturity/__tests__/maturity-summary.test.tsx
• Pre-push hooks: backend-tests and frontend-tests passed

---

Note

Medium Risk
Medium risk because it changes maturity analytics computation and API payload shape, which could affect dashboard correctness/performance and downstream consumers expecting the previous schema.

Overview
Adds harness_configuration_fingerprints to the maturity analytics response, computing a SHA-256–based fingerprint that groups sessions by agent_type, permission_mode, tool set, customization state/provenance, and presence/amount of context usage, and returns aggregate metrics (session/engineer counts, success + compound reliability, leverage, and “top” tools/customizations/signals).

Surfaces this new dataset in the frontend via a HarnessFingerprintTable on the Harness Intelligence “Customizations” tab (with empty state + tests), updates API types/schemas, and extends backend/frontend tests to cover the new response field and fingerprint calculations.

^{Reviewed by Cursor Bugbot for commit 3f679e2. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Context signal count is total, not average like peers
  • Changed context_signal_count from a raw sum accumulator to a list-and-average pattern matching tool_count and customization_count, so all three metrics are now per-session averages.

Or push these changes by commenting:

@cursor push 40a5268b6e

Preview (40a5268b6e)

diff --git a/src/primer/server/services/maturity_service.py b/src/primer/server/services/maturity_service.py
--- a/src/primer/server/services/maturity_service.py
+++ b/src/primer/server/services/maturity_service.py
@@ -1073,7 +1073,7 @@
                 "leverage_scores": [],
                 "tool_counts": [],
                 "customization_counts": [],
-                "context_signal_count": 0,
+                "context_signal_counts": [],
                 "tools": Counter(),
                 "customizations": Counter(),
                 "signals": Counter(),
@@ -1090,7 +1090,7 @@
                 bucket["success_sessions"].add(row.id)
         bucket["tool_counts"].append(len(tool_names))
         bucket["customization_counts"].append(len(customization_names))
-        bucket["context_signal_count"] += context_count
+        bucket["context_signal_counts"].append(context_count)
         bucket["tools"].update(tool_names)
         bucket["customizations"].update(customization_names)
         bucket["signals"].update(
@@ -1148,7 +1148,11 @@
             )
             if bucket["customization_counts"]
             else 0,
-            context_signal_count=bucket["context_signal_count"],
+            context_signal_count=round(
+                sum(bucket["context_signal_counts"]) / len(bucket["context_signal_counts"])
+            )
+            if bucket["context_signal_counts"]
+            else 0,
             top_tools=[tool for tool, _count in bucket["tools"].most_common(5)],
             top_customizations=[
                 customization for customization, _count in bucket["customizations"].most_common(5)

_{You can send follow-ups to the cloud agent here.}

^{Reviewed by Cursor Bugbot for commit 9a9aea3. Configure here.}