chore: ignore .claude/ and .claude.local.md

[howie]

Summary

• Add .claude/ and .claude.local.md to .gitignore

Changes

• .gitignore: two new entries to exclude Claude Code session files

These paths are generated by Claude Code and contain personal session state, worktrees, and local settings. They are developer-specific and should not be committed to the repository.

Test plan

• Verify .claude/ and .claude.local.md no longer appear in git status output after this change

Generated with Claude Code

[github-actions]

⚠️ This PR is missing a Discord Discussion URL in the body.

All PRs must reference a prior Discord discussion to ensure community alignment before implementation.

Please edit the PR description to include a link like:

Discord Discussion URL: https://discord.com/channels/...

This PR will be automatically closed in 3 days if the link is not added.

[shaun-agent]

OpenAB PR Screening

This is auto-generated by the OpenAB project-screening flow for context collection and reviewer handoff.
Click 👍 if you find this useful. Human review will be done within 24 hours. We appreciate your support and contribution 🙏

• Title: chore: ignore .claude/ and .claude.local.md
• Source: chore: ignore .claude/ and .claude.local.md #837
• Status: moved to PR-Screening
• Generated at: 2026-05-17T14:17:15.050Z
• Discord thread: not available

Screening report

posted screening comment: https://github.com//pull/837#issuecomment-4470947037 project action: moved item `PVTI_lADOEFbZWM4BUUALzgs-CAc` to `PR-Screening` in https://github.com/orgs/openabdev/projects/1

Intent

This PR is presented as repository hygiene: ignore Claude Code local/session artifacts so developer-specific state does not show up in git status or get committed accidentally. Operator-visible value is lower review noise and less risk of local agent state leaking into the repo.

The diff also includes unrelated changes to Cargo.lock and src/acp/pool.rs. That widens the real intent beyond .gitignore and should be treated as a scope problem before merge.

Feat

Classification: chore, with an unexpected refactor/fix-sized runtime change.

The stated behavior is simple: add .claude/ and .claude.local.md to .gitignore. The actual diff also changes ACP session pool locking from pool-wide mutable access to Arc<Mutex<AcpConnection>> per active connection, adjusts eviction/cleanup behavior around locked connections, and bumps the locked openab package version from 0.7.4 to 0.7.6.

Who It Serves

Primary beneficiary for the stated change: maintainers and contributors using Claude Code locally.

Primary beneficiary for the unstated src/acp/pool.rs change, if intentional: agent runtime operators, because it changes connection concurrency and eviction behavior.

Rewritten Prompt

For the .gitignore change:

Update the repository ignore rules so Claude Code local artifacts are not tracked. Add .claude/ and .claude.local.md to .gitignore. Verify that creating those paths locally does not appear in git status. Do not include unrelated Rust source or lockfile changes in this PR.

If the ACP pool change is intentional, split it into a separate PR with this brief:

Refactor SessionPool so the pool map lock is held only for connection lookup and each AcpConnection has its own async mutex. Preserve session resume, idle cleanup, eviction, and process cleanup semantics. Add concurrency-focused tests covering simultaneous use of different thread connections, cleanup skipping in-use connections, and eviction behavior when all connections are busy.

Merge Pitch

The ignore-rule change is low risk, useful, and mergeable as a tiny maintenance PR.

The likely reviewer concern is scope contamination: the PR title and body do not describe the ACP pool changes or the lockfile version bump. Those changes affect runtime behavior and deserve separate review, tests, and a different risk profile. As-is, this should not be merged without either removing the unrelated files or retitling/reworking the PR into a real concurrency change.

Best-Practice Comparison

OpenClaw and Hermes Agent are only lightly relevant to the .gitignore part; they do not materially inform ignore rules for local tool state.

They are relevant to the ACP pool diff. Both patterns favor explicit ownership of execution state, durable state boundaries, and predictable scheduling/session behavior. The proposed per-connection locking moves toward isolated executions by avoiding a single pool-wide lock during connection use, but it needs tests to prove it does not weaken eviction, cleanup, or delivery/session routing guarantees. Hermes-style self-contained runs and atomic persisted state are not directly implemented here.

Implementation Options

Option 1, conservative: keep only the .gitignore additions in this PR. Remove Cargo.lock and src/acp/pool.rs changes. Merge after a quick git status verification.

Option 2, balanced: split into two PRs. Merge the .gitignore PR quickly, then open a separate ACP pool concurrency PR with tests and a clear explanation of behavior changes.

Option 3, ambitious: convert this PR into a broader ACP pool concurrency and local-tooling hygiene PR, with title/body updated, tests added for locking/eviction/cleanup behavior, and reviewer attention explicitly requested for runtime semantics. This is slower and probably not worth mixing with ignore rules.

Comparison Table

Option	Speed	Complexity	Reliability	Maintainability	User Impact	Fit for OpenAB now
Conservative: .gitignore only	Fastest	Very low	High	High	Removes local noise	Best immediate fit
Balanced: split PRs	Fast for hygiene, moderate for pool work	Moderate	High if tests are added	High	Keeps both improvements reviewable	Best overall fit
Ambitious: keep combined scope	Slow	Higher	Unclear until tested	Lower due to mixed intent	Potential runtime benefit plus hygiene	Weak fit unless urgent

Recommendation

Use the balanced path. Ask the author to remove Cargo.lock and src/acp/pool.rs from this PR, merge the .gitignore change as a small chore, then resubmit the ACP pool changes separately with tests and an accurate title/body.

For routing: this item is ready for PR-Screening, but should not be advanced to implementation/review follow-up as a clean .gitignore PR until the unrelated runtime diff is split or explained.

[howie]

Discord Discussion URL: https://discord.com/channels/1491295327620169908/1491365158868619404/1505576771255533628