π The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
TypeScript agentic-aiagentic-frameworkagentic-workflowagentsai-agentsai-assistantai-skillsautonomous-agentsclaudecodexdsh-pluginharnessmcp-servermulti-agentmulti-agent-systemsnpmskillsswarmswarm-intelligencetypescript
perf(plugins): ADR-098 Part 2 (slice 3/4) β token diet for ruflo-adr/adr-architect
Reuven85eab480e9dd
diff --git a/plugins/ruflo-adr/.claude-plugin/plugin.json b/plugins/ruflo-adr/.claude-plugin/plugin.jsonindex 8bdbf6209b..a886ffb4fc 100644--- a/plugins/ruflo-adr/.claude-plugin/plugin.json+++ b/plugins/ruflo-adr/.claude-plugin/plugin.json@@ -1,9 +1,18 @@{"name": "ruflo-adr",- "description": "ADR lifecycle management β create, index, supersede, and link Architecture Decision Records to code",- "version": "0.1.0",- "author": { "name": "ruvnet", "url": "https://github.com/ruvnet" },+ "description": "ADR lifecycle management \u2014 create, index, supersede, and link Architecture Decision Records to code",+ "version": "0.1.1",+ "author": {+ "name": "ruvnet",+ "url": "https://github.com/ruvnet"+ },"homepage": "https://github.com/ruvnet/ruflo","license": "MIT",- "keywords": ["ruflo", "adr", "architecture", "decisions", "documentation"]+ "keywords": [+ "ruflo",+ "adr",+ "architecture",+ "decisions",+ "documentation"+ ]}diff --git a/plugins/ruflo-adr/REFERENCE.md b/plugins/ruflo-adr/REFERENCE.mdnew file mode 100644index 0000000000..b58178d81a--- /dev/null+++ b/plugins/ruflo-adr/REFERENCE.md@@ -0,0 +1,112 @@+# ruflo-adr β ADR Reference++Companion reference for `adr-architect`. The agent prompt deliberately stays lean per [ADR-098 Part 2](../../v3/docs/adr/ADR-098-plugin-capability-sync-and-optimization.md); this file collects the template, graph-storage recipes, and detection workflow the agent reads on-demand.++## ADR template++Every ADR follows this structure:++```markdown+# ADR-NNN: <Title>++- **Status**: proposed | accepted | deprecated | superseded by ADR-XXX+- **Date**: YYYY-MM-DD+- **Deciders**: <list of people>+- **Tags**: <comma-separated tags>++## Context++<What is the issue that we're seeing that motivates this decision?>++## Decision++<What is the change that we're proposing and/or doing?>++## Consequences++### Positive+- <good outcomes>++### Negative+- <trade-offs and costs>++### Neutral+- <other effects>++## Links+- Supersedes: ADR-XXX (if applicable)+- Amended by: ADR-YYY (if applicable)+- Related: ADR-ZZZ+```++## Lifecycle state machine++```+ (write)+ βΌ+ proposed+ β accept+ βΌ+ accepted βββdeprecateββββΊ deprecated+ β+ β supersede+ βΌ+ superseded by ADR-XXX+```++- An ADR enters `proposed` on creation.+- `accepted` after team review.+- `deprecated` when no longer recommended but kept for history.+- `superseded` is permanent; ADR-XXX must already exist and reference back via `supersedes:`.++## AgentDB graph storage++Persist the ADR tree + relationships so traversal queries (e.g. "all ADRs depended on by ADR-097") work without reparsing markdown:++```bash+# Hierarchical tree β store each ADR under adr/<id>+mcp__claude-flow__agentdb_hierarchical-store \+ --path "adr/ADR-097" \+ --value '{"status":"accepted","title":"Federation Budget Circuit Breaker","date":"2026-05-04"}'++# Causal edges for relationships+mcp__claude-flow__agentdb_causal-edge \+ --from "ADR-097" --to "ADR-086" --relation "depends-on"+mcp__claude-flow__agentdb_causal-edge \+ --from "ADR-098" --to "ADR-095" --relation "depends-on"+mcp__claude-flow__agentdb_causal-edge \+ --from "ADR-094" --to "ADR-093" --relation "amends"+```++Standard relationship vocabulary:+- `supersedes` β replaces an older ADR (the older one moves to `superseded`).+- `amends` β adjusts an existing ADR without superseding it (older stays `accepted`).+- `depends-on` β this ADR's correctness depends on another ADR's decision being in force.++## Code-ADR violation detection++Detect when code drifts away from an accepted ADR:++1. **Find ADR references in code** β `Grep` for the ADR ID in comments:+ ```bash+ grep -rnE "ADR-[0-9]{3}" src/ --include='*.ts' --include='*.js'+ ```+2. **Find when those code lines last changed** β `git blame` on each match.+3. **Compare against ADR status timestamps** β load ADR's `Date:` field and ADR-graph history. If the code changed *after* the ADR was accepted AND the ADR wasn't amended in the same PR, flag the file as a potential violation.+4. **Report** β emit a structured finding with file, line, ADR ID, and the lag in days between ADR acceptance and the code change.++Common drift patterns to look for:+- Code references a `superseded` ADR (should reference its successor).+- Code introduces a pattern explicitly forbidden in an `accepted` ADR.+- ADR's "Decision" wording doesn't match the actual code shape.++## Index file++`docs/adr/README.md` maintains a flat index of all ADRs with status. Update on every ADR write:++```markdown+| # | Title | Status | Date |+|---|---|---|---|+| ADR-097 | Federation Budget Circuit Breaker | accepted | 2026-05-04 |+| ADR-096 | Encryption at Rest | accepted | 2026-05-03 |+```diff --git a/plugins/ruflo-adr/agents/adr-architect.md b/plugins/ruflo-adr/agents/adr-architect.mdindex c850d57451..3a4cb34323 100644--- a/plugins/ruflo-adr/agents/adr-architect.md+++ b/plugins/ruflo-adr/agents/adr-architect.md@@ -6,80 +6,30 @@ model: sonnetYou are an Architecture Decision Record specialist. Your responsibilities:-1. **Create** new ADRs with sequential numbering (ADR-001, ADR-002...) in `docs/adr/`-2. **Maintain** the ADR lifecycle: proposed -> accepted -> deprecated -> superseded-3. **Link ADRs to code** via grep/git blame -- detect when code changes violate accepted ADRs-4. **Track relationships** between ADRs (supersedes, amends, depends-on)+1. **Create** new ADRs with sequential numbering (ADR-001, ADR-002 β¦) in `docs/adr/`.+2. **Maintain** the ADR lifecycle: `proposed` β `accepted` β `deprecated` β `superseded`.+3. **Link ADRs to code** via grep / git blame β detect when code changes violate accepted ADRs.+4. **Track relationships** between ADRs (`supersedes`, `amends`, `depends-on`).-### ADR Template+## Reference-Every ADR follows this structure:+The full ADR markdown template, the AgentDB graph-storage commands for persisting the ADR tree + relationships, and the code-ADR linking workflow live in [`REFERENCE.md`](../REFERENCE.md). Read it when you need an exact field, a hierarchical-store path, or the violation-detection grep pattern β keeping reference data out of the agent prompt costs ~40% fewer tokens per spawn (per ADR-098 Part 2).-```markdown-# ADR-NNN: <Title>+## Tools-- **Status**: proposed | accepted | deprecated | superseded by [ADR-XXX]-- **Date**: YYYY-MM-DD-- **Deciders**: <list of people>-- **Tags**: <comma-separated tags>+- `mcp__claude-flow__agentdb_hierarchical-store` / `agentdb_hierarchical-query` β ADR tree storage.+- `mcp__claude-flow__agentdb_causal-edge` / `agentdb_causal-query` β relationship tracking.+- `mcp__claude-flow__memory_store` / `memory_search` β semantic search.+- `Read`, `Write`, `Edit` β ADR file operations.+- `Grep`, `Glob` β code scanning.+- `Bash` β git operations (`blame`, `log`, `diff`).-## Context+## Cross-references-<What is the issue that we're seeing that motivates this decision?>+- **ruflo-jujutsu**: Use diff analysis on PRs to check ADR compliance before merge.+- **ruflo-docs**: Trigger doc generation when ADRs change status.-## Decision--<What is the change that we're proposing and/or doing?>--## Consequences--### Positive-- <good outcomes>--### Negative-- <trade-offs and costs>--### Neutral-- <other effects>--## Links-- Supersedes: ADR-XXX (if applicable)-- Amended by: ADR-YYY (if applicable)-- Related: ADR-ZZZ-```--### AgentDB Graph Storage--Store the ADR dependency graph using AgentDB:--- **Hierarchical store** for the ADR tree:- `mcp__claude-flow__agentdb_hierarchical-store` with path `adr/<adr-id>` and the ADR metadata as value-- **Causal edges** for supersedes/amends relationships:- `mcp__claude-flow__agentdb_causal-edge` with `from: <old-adr-id>`, `to: <new-adr-id>`, `relation: supersedes|amends|depends-on`--### Code-ADR Linking--Detect ADR violations by:-1. `Grep` for ADR references in code comments (e.g., `// ADR-042`, `# See ADR-042`)-2. `git blame` to find when ADR-referenced code was last changed-3. Compare change date against ADR status -- flag if code changed after ADR was accepted but ADR was not updated-4. Report violations with file paths, line numbers, and the relevant ADR--### Cross-References--- **ruflo-jujutsu**: Use diff analysis on PRs to check ADR compliance before merge-- **ruflo-docs**: Trigger doc generation when ADRs change status--### Tools--- `mcp__claude-flow__agentdb_hierarchical-store`, `mcp__claude-flow__agentdb_hierarchical-query` -- ADR tree storage-- `mcp__claude-flow__agentdb_causal-edge`, `mcp__claude-flow__agentdb_causal-query` -- relationship tracking-- `mcp__claude-flow__memory_store`, `mcp__claude-flow__memory_search` -- semantic search-- `Read`, `Write`, `Edit` -- ADR file operations-- `Grep`, `Glob` -- code scanning-- `Bash` -- git operations (blame, log, diff)--### Memory Learning+## MemoryStore ADR patterns and architectural decisions for cross-project learning:```bash@@ -87,10 +37,9 @@ npx @claude-flow/cli@latest memory store --namespace adr-patterns --key "decisionpx @claude-flow/cli@latest memory search --query "architectural decision" --namespace adr-patterns```-### Neural Learning+## Neural learning-After completing tasks, store successful patterns:+After completing tasks, feed the ADR-lifecycle learning so future ADR-violation detection compounds:```bashnpx @claude-flow/cli@latest hooks post-task --task-id "TASK_ID" --success true --train-neural true-npx @claude-flow/cli@latest memory search --query "ADR lifecycle patterns" --namespace patterns```