fmflurry/harness-manifesto

Custom OpenCode settings

This repository also configures its own agents. See what harness-manifesto tells them →

170Stars on the repository
203Mods indexed here, across every type
yesterdayLast push, which is what freshness is scored on
MITLicence, which decides whether bodies are shown

angular-cop

01

fmflurry/harness-manifesto

Agent Claude Code needs its repo

MUST delegate when user runs /cop-review or asks for pre-merge review of HEAD vs a target branch. Angular + TypeScript focused (signals, RxJS, clean architecture, flurryx, TS strict). Reads project AGENTS.md. Runs tsc + lint. Tiered output. Read-only — findings only.

not rated 170 yesterday A 70 tokens copy · 88% MIT

api-spec-architect

02

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for generating or updating OpenAPI 3.1 specifications and publishing them to SmartBear SwaggerHub. Produces one Zalando-compliant YAML per bounded context and optionally publishes via the SwaggerHub MCP.

not rated 170 yesterday A 48 tokens copy · 98% MIT

architect

03

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for architecture, system design, scalability, cross-module boundaries, or high-impact technical decisions.

not rated 170 yesterday A 23 tokens copy · 88% MIT

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate when build, typecheck, lint, or TypeScript errors occur. Fixes only those errors with minimal diffs.

not rated 170 yesterday C 32 tokens original MIT

code-reviewer

05

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate immediately after meaningful code changes or PR/pull request review requests. Reviews diffs for bugs, regressions, quality, maintainability, and tests. Read-only — findings only, no patches.

not rated 170 yesterday A 46 tokens original MIT

coder

06

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for any non-test implementation work: writing new code, applying spec, scaffolding modules, porting code, implementing fixes. Pure implementation only — caller owns planning, architecture, and review.

not rated 170 yesterday A 43 tokens original MIT

comment-judge

07

fmflurry/harness-manifesto

Agent Claude Code

Read-only comment judge. Applies rubric-based verdict to code comments: false/stale (Important), forbidden/narration/noise (Nit), unsure (❓). REVIEW mode for diffs; PURGE mode for directory scans. Emits JSON findings.

not rated 170 yesterday A 54 tokens copy · 91% MIT

conductor

08

fmflurry/harness-manifesto

Agent Claude Code

Primary orchestrator. Routes every task to the matching specialist subagent. Cannot write or edit files directly — all writes go through writer (docs) or coder (code). Launch with claude --agent conductor.

not rated 170 yesterday A 46 tokens copy · 86% MIT

database-reviewer

09

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for PostgreSQL, SQL, Supabase, RLS, migrations, schema design, query performance, or DB security. Read-only — findings only; repo changes go to coder.

not rated 170 yesterday A 43 tokens copy · 91% MIT

doc-updater

10

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for documentation, README updates, architecture notes, and codemaps. Keeps docs aligned with code changes.

not rated 170 yesterday A 28 tokens copy · 75% MIT

dotnet-cop

11

fmflurry/harness-manifesto

Agent Claude Code needs its repo

MUST delegate when user runs /cop-review or asks for pre-merge review of HEAD vs a target branch on a .NET project. .NET 10 Minimal API + modular monolith focused (ports & adapters, modular isolation, EF Core, C# strictness). Reads project AGENTS.md. Runs dotnet build + dotnet format --verify-no-changes. Tiered…

not rated 170 yesterday A 90 tokens original MIT

e2e-runner

12

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for browser automation, Playwright work, and critical user-flow E2E tests. Reads the project e2e override or discovered start command, ports, and test runner. Generates, runs, and reports artifacts.

not rated 170 yesterday A 52 tokens original MIT

ecosystem-auditor

13

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for /ecosystem-audit or explicit audits of the repository-controlled OpenCode/Claude Code ecosystem, skill/agent parity, or harness coherence. Read-only — produces one evidence-first report and makes no changes.

not rated 170 yesterday A 51 tokens copy · 88% MIT

gaudi

14

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for Blender and 3D art work: mesh modeling, scene composition, materials and shading, lighting and camera framing, rendering, rigging and animation, and glTF/three.js web export. Drives the Blender MCP server directly and consults the threejs skill for web delivery.

not rated 170 yesterday A 63 tokens copy · 94% MIT

gdpr-specialist

15

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for GDPR/privacy/data-protection compliance review of code (frontend, backend, database, infrastructure layers, full stack). FRANCE/CNIL-focused: audits against French GDPR rules, CNIL recommendations, Loi Informatique et Libertés. Covers consent flows, cookies/trackers, PII handling, data-subject…

not rated 170 yesterday A 124 tokens copy · 94% MIT

git-specialist

16

fmflurry/harness-manifesto

Agent Claude Code

Git workflow specialist. Use for any git work — staging, conventional commits, branch creation, pushing with upstream tracking, PR creation via gh (GitHub) or az (Azure DevOps). Auto-detects host from origin. Enforces strict commit and branch naming.

not rated 170 yesterday E 62 tokens original MIT

planner

18

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for complex feature planning, implementation plans, large refactors, or unclear execution order. Produces plans only.

not rated 170 yesterday A 27 tokens original MIT

postgres-dba

19

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for PostgreSQL instance operations: health monitoring, vacuum/bloat, WAL/checkpoints, backups/PITR, connection pooling, role/RLS runtime audit, container persistence & major-version upgrades. Advisory and read-only — emits human-confirmed commands, never executes mutations. NOT for SQL/migration/schema…

not rated 170 yesterday A 75 tokens copy · 94% MIT

refactor-cleaner

20

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for dead code, unused exports, duplication, and consolidation cleanup. Preserves behavior and verifies safety.

not rated 170 yesterday A 28 tokens original MIT

scout

21

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate when a dispatch needs a file list that isn't known yet. Scout emits a file manifest only — no analysis, no review, no recommendations. Hard cap 8 tool calls.

not rated 170 yesterday A 41 tokens copy · 92% MIT

security-reviewer

22

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for auth, user input, API endpoints, secrets, payments, PII, or sensitive data. Read-only — reports vulnerabilities. If the coder agent is installed, the orchestrator dispatches fixes to that agent. Otherwise the main agent applies the fixes.

not rated 170 yesterday A 58 tokens original MIT

tdd-guide

23

fmflurry/harness-manifesto

Agent Claude Code

MUST delegate for new features, bug fixes, or refactors that need tests. A refactor pins current behavior first with characterization tests that already pass. Enforces RED-GREEN-REFACTOR for new behavior and 80%+ coverage. Writes test files only; the orchestrator delegates implementation to coder.

not rated 170 yesterday A 67 tokens original MIT

why-investigator

24

fmflurry/harness-manifesto

Agent Claude Code

Spawned by the why skill only. Investigates one evidence category (git/gh or one MCP source) for design rationale. Read-only.

not rated 170 yesterday A 33 tokens copy · 100% MIT

At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: