feat: per-adapter sensitivity evidence line (issue #68)

docs/PRD.md

@@ -489,7 +489,7 @@ For the passed-on-arrival case, which occurred in 4 of 8 executed cycles in the
 | Command | Behaviour |
 |---|---|
 | `tdd sensitivity begin` | record `git diff` and the untracked-file set as the reference state |
-| `tdd sensitivity check` | run the suite with the agent's mutation in place; require the target to now fail; record the mutation diff and the observed failure |
+| `tdd sensitivity check` | run the suite with the agent's mutation in place; require the target to now fail; record the mutation diff and the observed failure; each adapter extracts a one-line `target_evidence` (the assertion line, not runner noise), stored in `sensitivity_check.evidence_line` (schema v7) and rendered as the `observed:` snippet in the friction log |
 | `tdd sensitivity end` | `git checkout --` the mutated tracked paths, then assert the resulting `git diff` is byte-identical to the reference; emit `restore_mismatch` on any difference |
 
 - **R8.4** A cycle that passed on arrival cannot reach `CLOSED` without a completed sensitivity
@@ -802,7 +802,7 @@ depend on any of them being installed.
 ### 13.1 Storage
 - **R13.1** SQLite, single file, append-only for invocations and events.
 - **R13.2** Schema versioned and migrated. The schema is the long-lived asset; the transport is not.
-  Current schema version: **3** (v3 adds the `advance_claim` table, R9.23).
+  Current schema version: **7** (v7 adds `sensitivity_check.evidence_line TEXT` for per-adapter assertion-line evidence; earlier milestones: v3 adds `advance_claim`, v4–v6 are intermediate columns).
 
 ### 13.2 Location
 - **R13.3** **One ledger per repository**, in a per-user data directory keyed by the repository's

src/tddcli/adapters/base.py

@@ -25,6 +25,7 @@ class Verdict:
     target: str | None = None
     target_outcome: str = NOT_FOUND
     target_failure: str = ""
+    target_evidence: str = ""
     passed: list[str] = field(default_factory=list)
     failed: list[str] = field(default_factory=list)
     duration_ms: int = 0

src/tddcli/adapters/exec_adapter.py

@@ -168,6 +168,9 @@ def run(self, target: str | None = None) -> Verdict:
                 if target == qualified:
                     verdict.target_outcome = FAILED
                     verdict.target_failure = clip_failure(combined)
+                    verdict.target_evidence = next(
+                        (ln for ln in reversed(combined.splitlines()) if ln.strip()), ""
+                    )
 
         verdict.duration_ms = int((time.monotonic() - started) * 1000)
         return verdict

src/tddcli/adapters/gradle_adapter.py

@@ -329,7 +329,11 @@ def run(self, target: str | None = None) -> Verdict:
             verdict.target_outcome = PASSED
         elif target in failed:
             verdict.target_outcome = FAILED
-            verdict.target_failure = failures.get(target, "test failed")
+            failure_text = failures.get(target, "test failed")
+            verdict.target_failure = failure_text
+            verdict.target_evidence = next(
+                (ln for ln in failure_text.splitlines() if ln.strip()), ""
+            )
         # else NOT_FOUND (default): the suite ran but never produced this id
 
         return verdict

src/tddcli/adapters/pytest_adapter.py

@@ -12,6 +12,7 @@
 from __future__ import annotations
 
 import json
+import re
 import shlex
 import tempfile
 from pathlib import Path
@@ -148,7 +149,9 @@ def run(self, target: str | None = None) -> Verdict:
         if hit is not None:
             verdict.target_outcome = PASSED if hit["outcome"] == "passed" else FAILED
             call = hit.get("call") or hit.get("setup") or {}
-            verdict.target_failure = clip_failure(str(call.get("longrepr", "")))
+            longrepr = str(call.get("longrepr", ""))
+            verdict.target_failure = clip_failure(longrepr)
+            verdict.target_evidence = self._evidence_line(longrepr)
         else:
             target_file = native.split("::", 1)[0]
             if any(c == target_file or c.startswith(target_file) for c in uncollectable):
@@ -160,6 +163,14 @@ def run(self, target: str | None = None) -> Verdict:
                 verdict.target_outcome = NOT_FOUND
         return verdict
 
+    @staticmethod
+    def _evidence_line(longrepr: str) -> str:
+        for line in longrepr.splitlines():
+            m = re.match(r"^E\s+(.*)", line)
+            if m:
+                return m.group(1)
+        return ""
+
     @staticmethod
     def _collector_error(collectors: list[dict], target_file: str) -> str:
         for collector in collectors:

src/tddcli/adapters/vitest_adapter.py

@@ -152,8 +152,13 @@ def run(self, target: str | None = None) -> Verdict:
                         self.normalise_id(self._id_for(suite.get("name", ""), t["fullName"]))
                         == ntarget
                     ):
+                        messages = t.get("failureMessages", [])
                         verdict.target_failure = "\n".join(
-                            clip_failure(m, 600) for m in t.get("failureMessages", [])[:3]
+                            clip_failure(m, 600) for m in messages[:3]
+                        )
+                        first_msg = messages[0] if messages else ""
+                        verdict.target_evidence = next(
+                            (ln for ln in first_msg.splitlines() if ln.strip()), ""
                         )
             return verdict
 

src/tddcli/adapters/xctest_adapter.py

@@ -272,7 +272,9 @@ def run(self, target: str | None = None) -> Verdict:
             verdict.target_outcome = PASSED
         elif target in verdict.failed:
             verdict.target_outcome = FAILED
-            verdict.target_failure = self._failure_for(combined, self.strip(target))
+            window = self._failure_for(combined, self.strip(target))
+            verdict.target_failure = window
+            verdict.target_evidence = self._evidence_line(window)
         # else NOT_FOUND (default)
 
         return verdict
@@ -289,6 +291,13 @@ def _build_errors(self, combined: str) -> str:
             if ": error:" in line or "** BUILD FAILED **" in line
         )
 
+    @staticmethod
+    def _evidence_line(window: str) -> str:
+        for line in window.splitlines():
+            if ": error:" in line:
+                return line
+        return ""
+
     def _failure_for(self, combined: str, native_id: str) -> str:
         """Capture the assertion lines between 'started' and 'failed' for one test.
 

src/tddcli/cli.py

@@ -1134,16 +1134,18 @@ def cmd_sensitivity(args) -> Envelope:
     projects = json.loads(cycle["projects"])
 
     if args.step == "check":
-        outcomes, _, _, failure_text = engine.run_projects(
+        outcomes, _, verdicts, failure_text = engine.run_projects(
             projects, targets, cycle, "SENSITIVITY", False
         )
         # A mutation that breaks collection also proves the test depends on the code.
         bites = bool(outcomes) and all(o in (FAILED, NOT_COLLECTED) for o in outcomes.values())
+        evidence = next((v.target_evidence for v in verdicts if v.target_evidence), "")
         ledger.update(
             "sensitivity_check",
             open_check["id"],
             mutation_diff=gitutil.diff_text(worktree)[:20000],
             observed_failure=failure_text[:4000],
+            evidence_line=evidence,
         )
         if not bites:
             return Envelope(

src/tddcli/ledger.py

@@ -13,7 +13,7 @@
 from datetime import datetime, timedelta, timezone
 from pathlib import Path
 
-SCHEMA_VERSION = 6
+SCHEMA_VERSION = 7
 
 
 class LedgerVersionError(RuntimeError):
@@ -37,6 +37,8 @@ class LedgerVersionError(RuntimeError):
     4: "ALTER TABLE baseline ADD COLUMN source TEXT NOT NULL DEFAULT 'probed';",
     # v5 -> v6 added ancillary_files column to plan_contract; ALTER TABLE covers old ledgers.
     5: "ALTER TABLE plan_contract ADD COLUMN ancillary_files TEXT NOT NULL DEFAULT '[]';",
+    # v6 -> v7 added evidence_line column to sensitivity_check; ALTER TABLE covers old ledgers.
+    6: "ALTER TABLE sensitivity_check ADD COLUMN evidence_line TEXT;",
 }
 
 SCHEMA = """
@@ -195,6 +197,7 @@ class LedgerVersionError(RuntimeError):
     reference_untracked TEXT NOT NULL,
     mutation_diff TEXT,
     observed_failure TEXT,
+    evidence_line TEXT,
     restored_ok INTEGER,
     opened_at TEXT NOT NULL,
     closed_at TEXT

src/tddcli/render.py

@@ -93,8 +93,15 @@ def friction_log(ledger: Ledger, run) -> str:
         if sens:
             a("- **Sensitivity check:** verified, restore byte-identical")
             if sens["observed_failure"]:
-                snippet = sens["observed_failure"].strip().splitlines()
-                a(f"  - observed: `{snippet[0][:160] if snippet else ''}`")
+                evidence = sens["evidence_line"]
+                if evidence:
+                    capped = ("…" + evidence[-160:]) if len(evidence) > 160 else evidence
+                    a(f"  - observed: `{capped}`")
+                elif evidence == "":
+                    a("  - observed: <no assertion line captured>")
+                else:
+                    snippet = sens["observed_failure"].strip().splitlines()
+                    a(f"  - observed: `{snippet[0][:160] if snippet else ''}`")
 
         commits = ledger.all(
             "SELECT * FROM commit_record WHERE cycle_id = ? ORDER BY id", (cycle["id"],)

tasks/friction-logs/issue-68-sensitivity-evidence-friction.md

@@ -0,0 +1,121 @@
+# Implementation Friction Log: tasks/issue-68-sensitivity-evidence.md
+
+- Run: 12
+- Executor: claude-sonnet-4-6 (source: transcript)
+- Plan blob: `34ede538eab3a01b75b9d9cdfde5c86000318fe0` (declared)
+- Started: 2026-08-28T16:04:38.652188+00:00  Ended: 2026-08-28T17:21:10.569424+00:00  Outcome: complete
+- Baseline failures at start: tddcli=0
+
+## Plan fidelity
+
+- Declared cycles: 11
+- Delivered: 11   Skipped: 0
+- Never reached: none
+- Human interventions: 0
+
+### Cycle 11: a long observed line is capped keeping the tail, not the head  _(standard)_
+- **Target:** `tddcli::tests/test_sensitivity_evidence.py::test_long_evidence_is_capped_keeping_the_tail`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `e6d5235ad` [red] test: an over-long observed line keeps its tail (1 files)
+  - `146e15314` [green] feat: tail-keeping cap on the observed evidence line (1 files)
+
+### Cycle 10: legacy rows with NULL evidence keep the first-line fallback  _(standard)_
+- **Target:** `tddcli::tests/test_sensitivity_evidence.py::test_null_evidence_falls_back_to_first_observed_line`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'SENSITIVITY': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** passed (**passed**)
+- **Sensitivity check:** verified, restore byte-identical
+  - observed: `repo = PosixPath('/private/var/folders/zl/3010c_557g5_2rm9tyqsc03h0000gp/T/pytest-of-headless-coding/pytest-239/test_null_evidence_falls_back_0/workspace')`
+- **Commits:**
+  - `a7004cc9f` [refactor] refactor: legacy rows with NULL evidence keep the first-line fallback (1 files)
+- **Event — red_first_violation:** ["tddcli::tests/test_sensitivity_evidence.py::test_null_evidence_falls_back_to_first_observed_line"]
+
+### Cycle 9: empty evidence renders an explicit no-assertion-line sentinel  _(standard)_
+- **Target:** `tddcli::tests/test_sensitivity_evidence.py::test_empty_evidence_renders_the_sentinel`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `8e68e09f0` [red] test: empty evidence renders <no assertion line captured> (1 files)
+  - `8e0308820` [green] feat: render the no-assertion-line sentinel instead of wire noise (1 files)
+
+### Cycle 8: friction log observed line renders the stored evidence line  _(standard)_
+- **Target:** `tddcli::tests/test_sensitivity_evidence.py::test_friction_log_observed_line_is_the_evidence_line`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `b2dc16f52` [red] test: observed line shows evidence_line, not the raw first line (1 files)
+  - `ddc60246f` [green] feat: friction log prefers evidence_line for the observed snippet (1 files)
+
+### Cycle 7: sensitivity check stores the adapter's evidence line in the ledger  _(standard)_
+- **Target:** `tddcli::tests/test_sensitivity_evidence.py::test_sensitivity_check_records_the_evidence_line`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 2}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `53ad9734d` [red] test: sensitivity check persists evidence_line on its ledger row (1 files)
+  - `619c0bd3d` [green] feat: schema v7 — sensitivity_check.evidence_line stored at check time (2 files)
+  - `739ed8a0f` [refactor] refactor: sensitivity check stores the adapter's evidence line in the ledger (1 files)
+
+### Cycle 6: exec evidence falls back to the last non-empty output line  _(standard)_
+- **Target:** `tddcli::tests/test_evidence_extraction.py::test_exec_evidence_is_the_last_nonempty_output_line`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 2}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `1df55d55d` [red] test: exec evidence is the last non-empty combined-output line (1 files)
+  - `78771ff01` [green] feat: exec evidence falls back to the last non-empty line (1 files)
+  - `9012b32fe` [refactor] refactor: exec evidence falls back to the last non-empty output line (1 files)
+
+### Cycle 5: gradle evidence is the first line of the junit failure message  _(standard)_
+- **Target:** `tddcli::tests/test_evidence_extraction.py::test_gradle_evidence_is_the_first_failure_message_line`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `e0a5301d1` [red] test: gradle evidence is the junit failure message line (1 files)
+  - `bebee75f9` [green] feat: gradle evidence extracted from the failure element's message (1 files)
+
+### Cycle 4: vitest evidence is the first line of the first failure message  _(standard)_
+- **Target:** `tddcli::tests/test_evidence_extraction.py::test_vitest_evidence_is_the_first_failure_message_line`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `3c0c01982` [red] test: vitest evidence is the first failureMessage line (1 files)
+  - `248e55040` [green] feat: vitest evidence extracted from failureMessages[0] (1 files)
+
+### Cycle 3: xctest evidence is the error line, not interleaved console noise  _(standard)_
+- **Target:** `tddcli::tests/test_evidence_extraction.py::test_xctest_evidence_is_the_error_line_not_console_noise`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 1, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 2}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `e04e0098a` [red] test: xctest evidence line ignores console noise in the test window (1 files)
+  - `5817c6ab2` [green] feat: xctest evidence is the first ': error:' line of the test's window (1 files)
+  - `1f1b40b52` [refactor] refactor: xctest evidence is the error line, not interleaved console noise (1 files)
+
+### Cycle 2: pytest evidence is empty when longrepr has no assertion line  _(pin)_
+- **Target:** `tddcli::tests/test_evidence_extraction.py::test_pytest_evidence_is_empty_when_no_assertion_line_exists`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_PIN': 1, 'SENSITIVITY': 1, 'CLOSE_SWEEP': 1}
+- **First run outcome:** passed (as expected)
+- **Sensitivity check:** verified, restore byte-identical
+  - observed: `tmp_path = PosixPath('/private/var/folders/zl/3010c_557g5_2rm9tyqsc03h0000gp/T/pytest-of-headless-coding/pytest-162/test_pytest_evidence_is_empty_0')`
+- **Commits:**
+  - `dad64c656` [pin] test: pin empty pytest evidence when no E-line exists (1 files)
+
+### Cycle 1: pytest evidence is the assertion line, not the xdist worker header  _(standard)_
+- **Target:** `tddcli::tests/test_evidence_extraction.py::test_pytest_evidence_is_the_assertion_line_not_the_xdist_header`
+- **Projects:** `tddcli`
+- **Suite runs by phase:** {'AWAITING_TEST': 2, 'AWAITING_IMPL': 1, 'CLOSE_SWEEP': 2}
+- **First run outcome:** failed (as expected)
+- **Commits:**
+  - `a4be0c744` [red] test: pytest evidence line skips the xdist worker header (1 files)
+  - `a2b7ecb47` [green] feat: Verdict.target_evidence — pytest extracts the first E-line of longrepr (2 files)
+  - `b464cf742` [refactor] refactor: pytest evidence is the assertion line, not the xdist worker header (1 files)
+