feat: project memory layer (tool + system context source)

[savagelysubtle]

Summary

Adds a Claude-style file-based memory layer so agents accumulate knowledge across sessions.

memory tool (packages/opencode/src/tool/memory.ts)

• Actions: list / read / write / delete, namespaced names supported (architecture/decisions)
• Storage: <xdg-data>/opencode/memory/<project-slug>/*.md — per-project isolation via a worktree slug
• Path traversal guarded; permission-gated like other tools

core/memory system context source

• Registers a refreshable context source that injects a memory index (file names + first meaningful line, capped at 50 entries) into the prompt baseline
• Uses the existing SystemContext algebra — memory updates render as context updates without restarting sessions

Design note

Follows Anthropic's memory-tool pattern: files curated by the agent itself rather than an embedding pipeline — just-in-time retrieval, works with compaction/context editing instead of against them.

Test plan

• New test/tool/memory.test.ts: write → list → read → delete round-trip; path traversal rejected
• Full test/tool/: 341 passing (1 pre-existing failure in write permissions test, reproduced on clean tree)
• bun typecheck clean in packages/core and packages/opencode

[github-actions]

This PR doesn't fully meet our contributing guidelines and PR template.

What needs to be fixed:

• PR description is missing required template sections. Please use the PR template.

Please edit this PR description to address the above within 2 hours, or it will be automatically closed.

If you believe this was flagged incorrectly, please let a maintainer know.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

Potential Duplicate PRs Found

Based on the search results, I found two related PRs that may be addressing similar memory functionality:

1. PR feat(opencode): native auto-memory for cross-session learning #20344: feat(opencode): native auto-memory for cross-session learning

   • feat(opencode): native auto-memory for cross-session learning #20344
   • Relevance: Directly related to cross-session knowledge accumulation, which is the core goal of this PR's memory layer.

2. PR feat(memory): implement persistent project memory and background extraction #43870: feat(memory): implement persistent project memory and background extraction

   • feat(memory): implement persistent project memory and background extraction #43870
   • Relevance: Appears to implement persistent project memory functionality, which overlaps significantly with the file-based memory layer being added in PR feat: project memory layer (tool + system context source) #44539.

Both of these PRs seem to address the same problem space of persistent memory across sessions. You should check their implementation details and status to see if they cover the same scope or if there's an opportunity to consolidate.

[github-actions]

This pull request has been automatically closed because it was not updated to meet our contributing guidelines within the 2-hour window.

Feel free to open a new pull request that follows our guidelines.