fix(pairing): add call-site verification to strategy-1 name matching (#14)

[jflowers]

Summary

Strategy-1 (name-convention) pairing in quality/pairing.py now verifies
call-site presence before emitting a pairing. When a test function's name
matches a target via prefix stripping (e.g., test_add → add), the
system walks the test function's AST body for direct calls to the matched
target. If no matching call is found, the pairing is suppressed entirely,
allowing the test to fall through to strategy-2 or strategy-3.

This eliminates false pairings where a test is named after a function it
never actually invokes, improving Gaze's test quality metrics accuracy.

Closes #14

How to Test

# Run the full test suite (574 tests, 95.78% coverage)
uv run pytest --cov=snake_eyes --cov-report=term-missing --cov-fail-under=85

# Run only the new call-site verification tests
uv run pytest tests/test_test_mapping_method.py -k "callsite" -v

# Run all CI gates
uv run ruff check src/ tests/
uv run ruff format --check src/ tests/
uv run mypy src/

How to Demo

1. Create a fixture project with a source file containing def add(): ...
   and a test file containing def test_add(): assert True (no call to add)
2. Run snake-eyes --stdio and send a test_mapping request
3. Observe that test_add does NOT pair with add at confidence 90
   (previously it would have been a false positive)
4. Add add() to the test body and re-run — now the pairing is emitted correctly

Key Files Changed

• src/snake_eyes/quality/pairing.py — Added optional test_tree parameter
  to _name_match; call-site verification via _direct_call_names before
  emitting pairings (both exact-match and case-only paths)
• tests/test_test_mapping_method.py — Updated 14 existing tests to include
  target calls; added 8 new tests covering call-site verification scenarios
• openspec/changes/fix-strategy1-callsite-verification/ — OpenSpec change
  artifacts (proposal, design, spec with 8 scenarios, tasks with 15 items)

This PR was generated by /uf.finale (AI-assisted).

[jflowers]

Council Verdict: APPROVE

Note: Could not post as APPROVE due to GitHub's self-review prohibition (author cannot approve own PR). Posted as COMMENT instead. Original verdict: APPROVE.

Reviewers: Adversary, Architect, Curator, Envoy, Guard, Herald, Scribe, SRE, Testing
Iterations: 1

Adversary (APPROVE)

No findings.

Architect (APPROVE)

2 LOW findings omitted.

Curator (APPROVE)

No findings.

Envoy (APPROVE)

3 LOW findings omitted.

Guard (APPROVE)

1 LOW finding omitted.

Herald (APPROVE)

3 LOW findings omitted.

Scribe (APPROVE)

5 LOW findings omitted.

SRE (APPROVE)

1 LOW finding omitted.

Testing (APPROVE)

3 LOW findings omitted.

---

This review was generated by /review-council (AI-assisted).

[jflowers]

PR Cost Report: #16

Session: Change proposal artifact workflow

ID: ses_fa122f139ffeQ63laznpIoOdU5

Metric	Value
Cost (parent only)	$11.12
Input tokens	119
Output tokens	51,384
Cache read tokens	9,707,717
Cache write tokens	797,731

Timeline: 2026-09-01 17:25:40 — 2026-09-02 08:33:07

Child sessions: 16 — additional cost: $15.58
Session tree total: $26.71

Child session breakdown

Session	Cost	Output Tokens
Adversary spec review (@divisor-adversary subagent)	$0.90	4,620
Architect spec review (@divisor-architect subagent)	$1.03	5,255
Guard spec review (@divisor-guard subagent)	$0.93	6,392
Testing spec review (@divisor-testing subagent)	$1.20	5,364
SRE spec review (@divisor-sre subagent)	$0.60	3,322
Spec re-review adversary (@divisor-adversary subagent)	$0.56	3,958
Spec re-review architect (@divisor-architect subagent)	$0.66	4,394
Spec re-review guard (@divisor-guard subagent)	$0.35	3,441
Spec re-review testing (@divisor-testing subagent)	$0.67	3,560
Spec re-review SRE (@divisor-sre subagent)	$0.50	2,667
Implement Group 1 + Group 2 tasks (@cobalt-crush-dev subagent)	$3.16	14,686
Code review - adversary (@divisor-adversary subagent)	$0.90	2,975
Code review - architect (@divisor-architect subagent)	$1.15	3,857
Code review - guard (@divisor-guard subagent)	$0.93	3,394
Code review - testing (@divisor-testing subagent)	$1.05	3,644
Code review - SRE (@divisor-sre subagent)	$1.00	2,917

---

Session: Code and spec review council

ID: ses_f9de40430ffeXowOdFqgj2VNfp

Metric	Value
Cost (parent only)	$4.10
Input tokens	59
Output tokens	24,227
Cache read tokens	3,175,236
Cache write tokens	304,546

Timeline: 2026-09-02 08:33:16 — 2026-09-02 09:27:47

Child sessions: 10 — additional cost: $12.23
Session tree total: $16.33

Child session breakdown

Session	Cost	Output Tokens
Gaze full quality report (@gaze-reporter subagent)	$2.27	10,549
Divisor Adversary review (@divisor-adversary subagent)	$1.13	3,176
Divisor Architect review (@divisor-architect subagent)	$1.24	4,450
Divisor Guard review (@divisor-guard subagent)	$1.22	4,197
Divisor Testing review (@divisor-testing subagent)	$1.17	4,061
Divisor SRE review (@divisor-sre subagent)	$1.17	4,046
Divisor Curator review (@divisor-curator subagent)	$1.05	3,223
Divisor Envoy review (@divisor-envoy subagent)	$1.08	2,834
Divisor Herald review (@divisor-herald subagent)	$0.96	2,453
Divisor Scribe review (@divisor-scribe subagent)	$0.95	3,008

---

Grand Total: $43.04 across 2 session tree(s) (28 sessions)