eclipse-tools

Autonomous bug hunter. Analyzes code, runs tests, navigates UI via agent-browser. Finds ONE concrete, reproducible bug per invocation. Returns structured JSON with repro steps, root cause (file:line), evidence, and proofs. Never fabricates evidence. Never fixes before reporting.

Reviews code quality WITHIN the scope contract only. Never suggests refactors outside allowed_paths. Checks: bugs, security, readability, test quality — scope-only. Anti drive-by enforcement.

Researches competitors via WebSearch. Finds alternatives, user complaints, market gaps. Returns structured JSON with traced sources and claims. Never fabricates URLs or data.

Analyzes codebase for extensible patterns, architecture opportunities, and feature additions. Read-only agent — never modifies code. Returns a strict JSON array of ideas.

Identifies code smells, duplication, dead code, naming issues, type safety gaps, and test coverage holes. Read-only agent. Returns a strict JSON array of ideas with severity.

Identifies documentation gaps: missing README sections, undocumented APIs, missing inline comments on complex logic. Read-only agent. Returns a strict JSON array of ideas.

Identifies performance optimization opportunities: N+1 queries, missing caching, bundle size, memory leaks, unoptimized rendering. Read-only agent. Returns a strict JSON array of ideas.

Identifies security vulnerabilities: OWASP Top 10, auth issues, input validation gaps, exposed secrets, dependency CVEs. Read-only agent. Returns a strict JSON array of ideas.

Identifies UI/UX and accessibility improvements via static code analysis. Browser live audit (via agent-browser) is handled by the ideation orchestrator skill, not this agent. Read-only agent. Returns a strict JSON array of ideas.

TDD implementer agent. Executes one task from a plan following strict RED-GREEN-COMMIT cycle. Produces TDD artifacts as proof including diffstat. Never touches files outside allowed_paths.

Analyzes a project to infer target audience, product vision, maturity, and competitive context. Strictly non-interactive — never asks questions. Returns a single JSON object.

Generates a prioritized feature roadmap with MoSCoW prioritization, phases, milestones, and testable acceptance criteria. Uses discovery and competitor context. Returns a single JSON object.

Reviews a completed task for spec compliance and TDD discipline. Checks: scope contract, TDD artifacts (including diffstat/lines_changed_per_file), definition of done. REFUSES if any check fails.

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.

Autonomous bug hunting skill. Explores code, tests, and UI (via agent-browser) to find concrete, reproducible bugs. Produces a bug packet with repro steps, root cause (file:line), and proof. Waits for explicit validation before fixing. After fix is validated and committed, automatically relaunches to hunt the next bug. Uses Mem0 (ECL-BUG) with heatmap strategy to explore least-covered zones first.

Execute a TDD implementation plan task-by-task using subagents. Each task: implementer (RED-GREEN-COMMIT) → spec-reviewer → code-quality-reviewer. Strict Mem0 transitions, max 3 review cycles, mandatory TDD artifacts.

Analyze competitors for the current project. Uses WebSearch to find alternatives, user complaints, and market gaps. Results stored in Mem0 with full source traceability. Generates derived ideas (ECL-IDEA) from high-opportunity gaps.

Brainstorm and design a solution for an accepted idea or feature. Explores the codebase, asks clarifying questions, proposes approaches, produces a design document with scope, acceptance criteria, and testing strategy. Updates Mem0 status to "designed".

Finalize a completed build: run full test suite, verify TDD artifacts (1 per task), prepare merge/PR text. Updates Mem0 status to "done".

Run a comprehensive ideation analysis on the current project. Launches 6 specialized agents in parallel to find code improvements, quality issues, documentation gaps, performance optimizations, security vulnerabilities, and UI/UX issues. Results are stored in Mem0 as draft ideas for triage.

Daily dashboard showing all ideas, features, bugs, and their status from Mem0. Read-only by default — only writes to Mem0 when user explicitly chooses an action. Use when starting a work session, checking progress, or deciding what to work on next.

Create a TDD implementation plan from a design document. Produces bite-sized tasks with exact file paths, test commands, allowed_paths scope contracts, and definitions of done. Detects the project's test runner and generates stack-specific commands. No code "novels" — each task is one RED-GREEN-COMMIT cycle.

Persist project insights, decisions, and workflow state into Mem0. Auto-invoked after ideation, roadmap, competitor, design, plan, build, finish, and bug-hunter workflows. Handles project_id scoping, schema versioning, deduplication, and migration.

Generate a product roadmap with MoSCoW-prioritized features, phases, and milestones. Optionally includes competitor analysis via --with-competitor flag. Results stored in Mem0 as ECL-FEAT items (not ECL-IDEA).