fix(attribution): scope non-agent file enumeration to parent commit

[peyton-alt]

Intermediate commits between session start and the attributed commit were inflating total_committed and deflating agent_percentage. The root cause: non-agent file detection used session_base → head as its window, so any file added in an unrelated commit appeared as a human addition.

Fix: pass HEAD's first parent tree/hash into CalculateAttributionWithAccumulated and use parent → head for non-agent file diffing. Only files that actually changed in this commit now count toward human additions. Falls back to session-base behaviour for initial commits (no parent).

Before/after on a 20-line agent commit with a 50-line unrelated intermediate commit:
total_committed: 70 -> 20
human_added: 50 -> 0
agent_percentage: 28.6% -> 100.0%

Changes:

• CalculateAttributionWithAccumulated: new parentTree/parentCommitHash params
• condenseOpts/attributionOpts: carry parent context through post-commit path
• PostCommit hook: resolve and thread first-parent tree/hash at commit time
• Tests: unit regressions, integration regression, Bug B repro (t.Skip)

---

Note

Medium Risk
Changes attribution baselining logic used during manual-commit condensation, which can affect reported agent/human line counts and percentages across commits. Although it’s metrics-only, it touches core post-commit/condensation paths and adds new git-object resolution fallbacks.

Overview
Fixes manual-commit attribution inflation by scoping non-agent file detection to the attributed commit only (diffing parent→HEAD instead of session base→HEAD), preventing unrelated intermediate commits from counting as human additions.

Threads the HEAD first-parent tree/hash through the post-commit and condensation pipeline into CalculateAttributionWithAccumulated, with best-effort resolution and warnings/fallbacks when parent context can’t be loaded.

Adds unit and integration regressions covering the intermediate-commit case, plus a focused (skipped) repro test for an outstanding “0% AI” investigation.

^{Written by Cursor Bugbot for commit e09d085. Configure here.}

[Copilot]

Pull request overview

This PR fixes attribution inflation caused by counting non-agent file changes across the entire session window (session base → HEAD) instead of scoping non-agent diffs to the attributed commit (parent → HEAD). This improves total_committed, human_added, and agent_percentage accuracy when intermediate unrelated commits exist.

Changes:

• Thread HEAD first-parent context (tree + hash) through the post-commit condensation path into attribution calculation.
• Update CalculateAttributionWithAccumulated to diff non-agent files using parent → head (falling back to session base behavior when no parent exists).
• Add unit + integration regressions, plus focused “Bug B” repro candidates (one intentionally skipped).

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
cmd/entire/cli/strategy/manual_commit_attribution.go	Uses parent→head baseline for non-agent file enumeration/diffing in attribution.
cmd/entire/cli/strategy/manual_commit_condensation.go	Extends condense/attribution opts to carry parent tree/hash into attribution.
cmd/entire/cli/strategy/manual_commit_hooks.go	Resolves HEAD parent tree and threads parent hash into condensation.
cmd/entire/cli/strategy/manual_commit_attribution_test.go	Adds unit tests validating parent-baseline behavior + fallback when parent is nil.
cmd/entire/cli/strategy/manual_commit_test.go	Adds Bug B candidate repro tests (one skipped).
cmd/entire/cli/integration_test/attribution_test.go	Adds integration regression for intermediate unrelated commit not inflating attribution.

[peyton-alt]

@BugBot review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR