Phase 19: OpenCode Commands and Skills
- OpenCode skill format (SKILL.md with YAML frontmatter)
- OpenCode command format ($ARGUMENTS substitution)
- OpenCode agent format (mode, tools, permissions)
- Directory structure for .opencode/ plugin
- Pitfalls for trigger patterns and argument handling
- README requirements for installation guide

Phase 19: OpenCode Commands and Skills
- 5 plans in 2 waves
- 3 parallel (commands, core skills, teleport skills)
- 2 sequential (agent, README)
- Ready for execution

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

…n preservation for R1.3.4

- Port Claude Code memory-search to OpenCode format
- Replace parameters: array with $ARGUMENTS documentation
- Add Arguments section with positional and flag parsing
- Reference memory-query skill
- Override global gitignore for .opencode directories

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Copy SKILL.md with YAML frontmatter (name, description, license, metadata)
- Copy references/command-reference.md with full CLI reference
- Skill format is portable between Claude Code and OpenCode

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Port Claude Code memory-recent to OpenCode format
- Replace parameters: array with $ARGUMENTS documentation
- Add Arguments section with --days and --limit flags
- Reference memory-query skill

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Copy SKILL.md with tier detection and intent classification guidance
- Copy references/command-reference.md with gRPC and CLI reference
- Portable skill format requires no modifications

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Copy SKILL.md with topic exploration and lifecycle documentation
- Copy references/command-reference.md with gRPC and CLI reference
- Portable skill format requires no modifications

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Port Claude Code memory-context to OpenCode format
- Replace parameters: array with $ARGUMENTS documentation
- Add Arguments section with grip ID and --before/--after flags
- Include error handling table
- Reference memory-query skill

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Copy SKILL.md with YAML frontmatter (name, description, license, metadata)
- Copy references/command-reference.md with full CLI reference
- Skill format is portable between Claude Code and OpenCode

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- SUMMARY.md with execution metrics and self-check
- STATE.md updated to Phase 19, Plan 2 of 5 complete

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Copy SKILL.md with YAML frontmatter (name, description, license, metadata)
- Copy references/command-reference.md with full CLI reference
- Includes vector-search, hybrid-search, and vector-stats commands

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add 19-01-SUMMARY.md with execution results
- 3 tasks completed: memory-search, memory-recent, memory-context commands
- 1 deviation: gitignore fix for .opencode directories

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- SUMMARY.md with execution metrics and self-check
- STATE.md updated to 3/5 plans complete

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…e guide

- Document global and per-project installation methods
- Document all three commands with usage examples and arguments
- Document @memory-navigator agent invocation with intent examples
- Document retrieval tiers, skills, architecture, and troubleshooting

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add memory-navigator.md with OpenCode-specific frontmatter (mode: subagent, tools, permission)
- Remove triggers: field (not supported by OpenCode), document patterns in body instead
- Preserve complete Process section with workflow, intent classification, tier detection (R1.3.4)
- Add Trigger Patterns section documenting when to invoke @memory-navigator
- Reference all five skills in body content (memory-query, topic-graph, bm25-search, vector-search, retrieval-policy)
- Restrict bash permissions to memory-daemon and grep commands only

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Exclude OS files, editor artifacts, local dev files, and build output

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Create 19-04-SUMMARY.md with execution results and self-check
- Update STATE.md to 4 of 5 plans complete

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Create 19-05-SUMMARY.md with execution results
- Update STATE.md: Phase 19 complete (5/5 plans), progress 33% (2/6 phases)
- Unblock Phase 20 (OpenCode Event Capture + Unified Queries)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

… contradicted by conditional action logic

…ough ingest

- Add agent: Option<String> to HookEvent with with_agent() builder
- Add agent: Option<String> to CchEvent with serde(default)
- Propagate agent through map_cch_to_hook() -> map_hook_event() -> Event.with_agent()
- Add tests: map_with_agent, map_without_agent, parse_with_agent, backward_compat, end_to_end
- All 27 tests passing across memory-client and memory-ingest, zero clippy warnings

- Replace hardcoded agent: None with r.metadata.get("agent").cloned()
- Add test_retrieval_result_agent_from_metadata with present/absent cases
- All 64 memory-service tests passing, zero clippy warnings

- SUMMARY.md with self-check PASSED
- STATE.md updated: Phase 20 Plan 1/3, key decisions recorded

- Add memory-capture.ts with four lifecycle hooks: session.created,
  session.idle, message.updated, tool.execute.after
- Tag all events with agent:opencode and project directory context
- Use fail-open pattern (try/catch) to never block OpenCode
- Defensive session ID extraction handles multiple input shapes
- Session idle maps to Stop event for R1.4.1/R1.4.2

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add 20-02-SUMMARY.md with execution results
- Update STATE.md: plan 2/3 complete, decisions, progress

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Create docs/adapters/clod-format.md with full CLOD specification (2 examples)
- Add crates/memory-daemon/src/clod.rs module with parse/validate/generate
- Four generator functions: claude, opencode, gemini, copilot
- CLI supports `clod convert --input --target --out` and `clod validate`
- Add toml dependency to workspace and memory-daemon
- 11 new tests for CLOD parser, generators, and CLI args

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…regation

- Add optional agent_filter field (201) to GetTopTopicsRequest in proto
- Implement get_topics_for_agent() in TopicStorage using TopicLink -> TocNode -> contributing_agents path
- Wire agent filter in TopicGraphHandler.get_top_topics() RPC handler
- Add get_top_topics_for_agent() to memory-client for agent-filtered queries
- TopicGraphHandler now accepts main_storage for TocNode lookups
- Backward compatible: omitting agent_filter returns all topics as before

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Create docs/adapters/cross-agent-guide.md (319 lines) covering all four adapters,
  agent discovery commands, filtered queries, and common workflows
- Create docs/adapters/authoring-guide.md (582 lines) covering AgentAdapter trait,
  event capture (hooks, session ID, fail-open, redaction, ANSI stripping),
  skills format, commands, agent tagging, config precedence, and testing

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…ueries

- Add Topics variant to AgentsCommand enum with --agent, --limit, --addr flags
- Implement agents_topics() function with formatted table output
- Handle empty results with informative message
- Wire Topics variant in handle_agents_command dispatch
- Add 2 CLI parse tests for agents topics (defaults and all options)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add Supported Agents table to docs/README.md with all four adapters
- Add Cross-Agent Discovery section with CLI examples
- Add documentation links for cross-agent guide, authoring guide, CLOD spec
- Add v2.2.0 Multi-Agent Ecosystem section to docs/UPGRADING.md
- Document new CLI commands, migration steps, and verification procedure
- Full workspace clippy and doc build pass

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Create 23-03-SUMMARY.md with execution metrics and decisions
- Update STATE.md: Phase 23 complete (3/3), v2.1 milestone at 100%

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- 5 unit tests for get_topics_for_agent: matching, no match, limit, sort order, case-insensitive
- 3 integration tests for GetTopTopics handler: without filter, with agent filter, empty filter
- All 8 new tests verify backward compatibility (no filter = all topics)
- Full workspace test, clippy, and doc pass

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add 23-02-SUMMARY.md with execution details, decisions, and metrics
- Update STATE.md with Phase 23 Plan 02 decisions

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Archive milestone roadmap and requirements to milestones/v2.1-*.
Update PROJECT.md with v2.1 validated requirements and key decisions.
Reorganize ROADMAP.md with collapsed milestone summaries.
Clean STATE.md for next milestone planning.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Replace stub get_ranking_status with actual SalienceConfig/NoveltyConfig defaults
- salience_enabled=true, usage_decay_enabled=true, novelty_enabled=false
- vector_lifecycle_enabled reflects whether vector service is configured
- Add test_get_ranking_status_returns_defaults unit test
- Fix blocking compile errors: add agent field to VectorMatch, TeleportSearchResult,
  and RrfEntry structs (Rule 3 - proto changes were uncommitted on branch)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add count_sessions_per_agent() method scanning events for distinct session_ids
- Bounded to last 365 days for performance
- Merge session counts into agent summaries from TOC aggregation
- Update test_list_agents_aggregates_from_toc_nodes to verify session_count=0
- Add test_list_agents_session_count_from_events verifying distinct session counting

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add optional string agent = 6 to TeleportSearchResult proto message
- Add optional string agent = 6 to VectorMatch proto message
- Add agent field to SearchSchema with STRING | STORED indexing
- Add agent field to TeleportResult and extract from BM25 documents
- Index agent from TocNode.contributing_agents in document builder
- Add agent field to VectorEntry with serde(default) for backward compat
- Add with_agent() builder method on VectorEntry

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Map agent field from TeleportResult to TeleportSearchResult proto
- Map agent field from VectorEntry to VectorMatch proto
- Add agent field to VectorSearchResult for retrieval handler
- Add test_handle_teleport_search_with_agent test (contributing_agents)
- Add test_handle_teleport_search_without_agent test (None case)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add 24-02-SUMMARY.md with execution results
- Update STATE.md: plan 2/3, 67% progress, tech debt resolved

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Replace stub with real implementation that reads vector metadata entries
- Filter entries by level (segment/grip/day/week) and retention cutoff
- Support dry_run mode, level filtering, and age_days_override
- Protect month/year levels from pruning
- Add metadata() accessor to VectorTeleportHandler
- Add test_prune_vector_index_no_service unit test

- Replace stub with report-only implementation scanning indexed documents
- Add count_docs_before_cutoff method to TeleportSearcher for level/timestamp analysis
- Support dry_run mode, level filtering, and age_days_override
- Protect month/year levels from pruning
- Report eligible document counts per level with actionable message
- Add test_prune_bm25_index_no_service unit test

- PruneVectorIndex and PruneBm25Index stubs resolved
- All 3 plans in Phase 24 (Proto & Service Debt Cleanup) complete
- STATE.md updated with decisions and phase completion

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Add 25-CONTEXT.md with implementation decisions for E2E test suite:
dual-location tests (e2e crate + daemon), structural+content assertions,
full provenance chain verification, pretty_assertions, ordering checks.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>