fix: store tree hash in checkpoint metadata for linkage resilience

[peyton-alt]

Summary

Store the commit's tree hash in checkpoint metadata during condensation. This is the CLI-side of the fix for #834.

Problem (#834)

When users rewrite git history (git rebase -i reword, git filter-branch), the Entire-Checkpoint trailer is stripped from the commit message. The checkpoint data still exists on entire/checkpoints/v1, but the rewritten commit no longer points to it. The user loses their agent label and session link on entire.io.

How tree hash fixes this

A commit's tree hash represents its code content. History rewrites change the commit SHA and can strip the message, but the tree hash stays the same (same code = same tree). By storing the tree hash during condensation, the web-side can match rewritten commits to their original checkpoints via tree hash when no trailer is found.

Example: User commits with trailer → checkpoint stored with tree_hash: abc123 → user runs git rebase -i reword → new commit has different SHA, no trailer, but same tree hash abc123 → web webhook matches tree hash → linkage restored.

What's in this PR (CLI-side)

• Add tree_hash field to CommittedMetadata and WriteCommittedOptions
• Populate tree hash from the HEAD commit during PostCommit condensation
• Tree hash is written to metadata.json on entire/checkpoints/v1

What's in the companion PR (web-side, entireio/entire.io)

• DB migration: tree_hash column on repo_checkpoints with index
• Webhook: fallback tree hash lookup for commits without trailers
• Webhook: lazy backfill tree hash for existing checkpoints

Also in this PR

Agent-initiated git revert/git cherry-pick now gets a checkpoint trailer when an agent session is ACTIVE. Previously prepare-commit-msg unconditionally skipped during sequence operations. This is a separate improvement, not part of the #834 fix.

Testing

• TestWriteCommitted_IncludesTreeHash — tree hash stored and read back from metadata
• TestShadowStrategy_PrepareCommitMsg_AgentRevertGetsTrailer — agent revert gets trailer
• TestShadowStrategy_PrepareCommitMsg_UserRevertSkipped — user revert still skipped
• Full CI passes (unit + integration + E2E canary)

🤖 Generated with Claude Code

---

Note

Medium Risk
Adds a new persisted tree_hash field to committed checkpoint metadata and changes git hook behavior during sequence operations, which could affect checkpoint linkage and when trailers are added to commits.

Overview
Improves checkpoint linkage resilience by persisting the condensed commit’s git tree hash (tree_hash) into session-level committed metadata (WriteCommittedOptions/CommittedMetadata) and wiring it through manual-commit condensation so rewritten commits can be re-associated even if the Entire-Checkpoint trailer is lost.

Adjusts manual-commit prepare-commit-msg behavior so rebase/cherry-pick/revert operations are only skipped when there is no ACTIVE agent session in the worktree; agent-initiated reverts now get checkpoint trailers while user-initiated ones remain unchanged. Adds focused tests covering tree_hash round-trip persistence and the new revert trailer behavior.

^{Written by Cursor Bugbot for commit 9164d8e. Configure here.}

[Copilot]

Pull request overview

Stores a commit’s git tree hash in checkpoint metadata during condensation so the web side can re-link checkpoints after history rewrites that drop the Entire-Checkpoint trailer, and tweaks prepare-commit-msg behavior so agent-driven revert/cherry-pick operations still get checkpoint trailers when a session is active.

Changes:

• Add tree_hash to committed checkpoint metadata (CommittedMetadata, WriteCommittedOptions) and populate it during PostCommit condensation from the HEAD commit’s tree hash.
• Allow prepare-commit-msg to proceed during git sequence operations (revert/cherry-pick/rebase) when an ACTIVE session exists in the current worktree; keep skipping when no active session.
• Add unit tests for tree_hash persistence and the new revert trailer behavior split by agent-active vs user/manual.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
cmd/entire/cli/strategy/manual_commit_test.go	Adds tests ensuring agent revert gets a trailer when session is ACTIVE and user revert is skipped when not active.
cmd/entire/cli/strategy/manual_commit_session.go	Adds helper to detect whether any session in the current worktree is ACTIVE.
cmd/entire/cli/strategy/manual_commit_hooks.go	Adjusts PrepareCommitMsg sequence-operation skip logic; passes commit tree hash into condensation options.
cmd/entire/cli/strategy/manual_commit_condensation.go	Threads treeHash through condensation options into committed checkpoint write options.
cmd/entire/cli/checkpoint/committed.go	Writes TreeHash into per-session committed metadata.json.
cmd/entire/cli/checkpoint/checkpoint.go	Extends checkpoint option/metadata structs to include TreeHash serialized as tree_hash.
cmd/entire/cli/checkpoint/checkpoint_test.go	Adds coverage that tree_hash is written and read back from committed metadata.

[Copilot]

In this test, metaDir is built with filepath.Join and then passed as StepContext.MetadataDir. MetadataDir is a repo-relative path used for git tree entries/commit trailers, so it needs forward slashes regardless of OS. Using filepath.Join will produce backslashes on Windows, which can break shadow-branch metadata paths and make the test (and any code paths it exercises) behave differently across platforms. Consider keeping MetadataDir as a slash-separated string (e.g., ".entire/metadata/") and deriving an absolute filesystem path separately (e.g., filepath.FromSlash + filepath.Join).

Suggested change

	metaDir := filepath.Join(".entire", "metadata", "agent-revert-session")
	require.NoError(t, os.MkdirAll(filepath.Join(dir, metaDir), 0o755))
	transcript := `{"type":"human","message":{"content":"revert the change"}}` + "\n" +
	`{"type":"assistant","message":{"content":"I'll revert that"}}` + "\n"
	require.NoError(t, os.WriteFile(filepath.Join(dir, metaDir, "full.jsonl"), []byte(transcript), 0o644))
	metaDir := ".entire/metadata/agent-revert-session"
	metaDirFS := filepath.Join(dir, filepath.FromSlash(metaDir))
	require.NoError(t, os.MkdirAll(metaDirFS, 0o755))
	transcript := `{"type":"human","message":{"content":"revert the change"}}` + "\n" +
	`{"type":"assistant","message":{"content":"I'll revert that"}}` + "\n"
	require.NoError(t, os.WriteFile(filepath.Join(metaDirFS, "full.jsonl"), []byte(transcript), 0o644))

[cursor]

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

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

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

[cursor]

Agent revert trailer added but condensation skipped in PostCommit

Medium Severity

PrepareCommitMsg now adds a checkpoint trailer during agent-initiated revert/cherry-pick (active session + sequence operation), but the PostCommit handler still sets IsRebaseInProgress: true for all sequence operations via isGitSequenceOperation(ctx). Since REVERT_HEAD/CHERRY_PICK_HEAD remain present during the post-commit hook, the state machine receives IsRebaseInProgress: true and emits no ActionCondense, so no checkpoint data is ever written. The commit ends up with a dangling Entire-Checkpoint trailer pointing to a nonexistent checkpoint ID.

Additional Locations (1)

• cmd/entire/cli/strategy/manual_commit_hooks.go#L874-L878