feat(tooling): add Claude harness for protocol code review and development#2277
Open
nimrod-teich wants to merge 1 commit intomainfrom
Open
feat(tooling): add Claude harness for protocol code review and development#2277nimrod-teich wants to merge 1 commit intomainfrom
nimrod-teich wants to merge 1 commit intomainfrom
Conversation
…pment Adds a self-contained agent team harness under .claude/ that structures multi-angle review and design-first development for the Lava off-chain protocol layer (rpcconsumer / rpcprovider / rpcsmartrouter / chainlib / lavasession / relaycore / statetracker / chaintracker / provideroptimizer / lavaprotocol / qos / metrics). Scope explicitly excludes on-chain x/ Cosmos modules. Two orchestrator skills: - protocol-review: 4 parallel reviewers (architecture, security, performance, style) fan out, then synthesis-reporter merges findings into a single prioritized report with deduplication, severity reconciliation, clustering, and effort estimates. - protocol-develop: architect -> implementer -> tester pipeline, with optional auto-triggered post-implementation review for changes touching signing paths, lavasession, rewardserver, or >3 packages. 8 agent definitions + 8 agent-level methodology skills + 3 shared references (protocol-map, review-taxonomy, report-template) ensure stable finding IDs across runs, consistent severity scale, and predictable merge output. CLAUDE.md registers the harness (agent table, skill table, execution rules, directory map, change log) so new sessions activate it automatically. Notes: - Removes .claude/ from .gitignore so the shared harness lives in the repo; local per-developer overrides (.claude/settings.local.json, .claude/*.local.*) remain ignored. - Adds _workspace/ (runtime artifact dir) and REVIEW_REPORT.md to .gitignore. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Codecov Report✅ All modified and coverable lines are covered by tests.
Flags with carried forward coverage won't be shown. Click here to find out more. 🚀 New features to boost your workflow:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a structured Claude Code harness under
.claude/for multi-angle review and design-first development of the Lava off-chain protocol layer. Review output is a single prioritized report merged from 4 parallel specialists (architecture, security, performance, style). Development runs as an architect → implementer → tester pipeline with optional auto-triggered post-implementation review.Scope boundary:
protocol/**only. On-chainx/Cosmos modules,app/, and consensus code are explicitly excluded by every agent's role definition and by the orchestrator scope rules.What's included
2 orchestrator skills:
protocol-review— parallel fan-out (4 reviewers) → synthesis-reporter fan-in →REVIEW_REPORT.mdprotocol-develop— pipeline (design → implement → test), with optional post-review for changes touching signing paths,lavasession,rewardserver, or >3 packages8 agents (role definitions + methodology skills):
Shared references:
protocol-map.md— package topography for all reviewers (in scope / out of scope / hot path)review-taxonomy.md— severity scale, category vocabularies, stable finding IDs (ARCH-NNN,SEC-NNN,PERF-NNN,STYLE-NNN), machine-readable JSON schemareport-template.md— synthesis output template with prioritized top findings + clusters + disagreements appendixCLAUDE.md registers the harness for new sessions: agent table, skill table, execution rules, directory map, workspace layout, change log.
Notable changes
.gitignore— removes.claude/(so the shared harness lives in the repo) and adds:.claude/settings.local.json,.claude/*.local.*) stay ignored_workspace/,_workspace_*/,REVIEW_REPORT.md) are ignoredCLAUDE.md— was previously untracked; now committed with harness registration section appended to the existing project overview. Existing content (build, testing, architecture, dependencies) is preserved.How it gets used
protocol/rpcprovider/" → triggersprotocol-review→ spawns the 4 reviewers in parallel via Agent tool withmodel: opus→ synthesis-reporter producesREVIEW_REPORT.mdwith severity counts, top findings, clusters, and fix order.protocol-develop→ architect produces design doc → implementer writes Go → tester adds race-clean tests → optional post-review since it touches hot-path and lavasession.Test plan
git grep -l ".claude/agents" .claude/skills/resolves cleanly (cross-refs valid)head -5 .claude/agents/*.md .claude/skills/*/SKILL.mdshows valid YAML frontmatter (name + description) on every filefind .claude -name SKILL.md -exec wc -l {} +shows all SKILL.md bodies under the 500-line progressive-disclosure budgetprotocol/rpcsmartrouter/last 3 commits) and verify:_workspace/01_*_findings.md_workspace/02_synthesis_report.md+REVIEW_REPORT.md_workspace/01_design.md_workspace/02_implementation.md+make lintclean_workspace/03_testing.md+-racecleanFollow-ups (out of scope for this PR)
🤖 Generated with Claude Code