Skip to content

AGENTS-INDEX.md

Latest commit

 

History

History
95 lines (82 loc) · 3.7 KB

AGENTS-INDEX.md

File metadata and controls

95 lines (82 loc) · 3.7 KB

Codebase Context Skill — Agents Index v2.0.0

For orchestrators: Quick reference for all available agents. Read agents/<name>.md for full instructions.

Agent Registry

# Agent Model Role Invocation
1 context-builder opus Deep codebase scan → .ccs/ index files After init or refresh
2 test-runner sonnet Execute tests, track results, auto-fix After test or fix
3 code-auditor opus Security, perf, dead code, a11y analysis After audit
4 git-tracker sonnet Branch, PR, merge, diff, sync ops After any git skill
5 knowledge-guide haiku Methodology guidance, note quality Advisory — any time
6 team-lead opus Multi-agent coordination, task decomposition After team

Tool Matrix

Agent Read Write Edit Glob Grep Bash Task
context-builder x x x x
test-runner x x x x
code-auditor x x x
git-tracker x x x x
knowledge-guide x x
team-lead x x x x x x

Agent → Skill Mapping

Agent Primary Skills When to Invoke
context-builder init, refresh New codebase, stale index, major changes
test-runner test, fix Test execution, failure diagnosis, auto-fix
code-auditor audit, review Security review, performance check, code quality
git-tracker branch, pr, merge, diff, sync, log, stash Any git workflow operation
knowledge-guide — Proactive guidance on methodology, note quality
team-lead team Complex multi-part tasks requiring parallel agents

Parallel Safety

Agent Can Run In Parallel Writes To
context-builder no .ccs/*.md
test-runner yes (per test suite) .ccs/task.md (append)
code-auditor yes stdout only
git-tracker no .ccs/branches/.md, .ccs/pulls/.md
knowledge-guide yes stdout only
team-lead no .ccs/team-board.md, .ccs/task.md

For SDK Integration

from claude_agent_sdk import AgentDefinition

agents = {
    "context-builder": AgentDefinition(
        description="Deep codebase analysis, generates .ccs/ index files",
        prompt=open("agents/context-builder.md").read(),
        tools=["Read", "Glob", "Grep", "Write"],
        model="opus"
    ),
    "test-runner": AgentDefinition(
        description="Run tests, track results, auto-fix failures",
        prompt=open("agents/test-runner.md").read(),
        tools=["Bash", "Read", "Grep", "Write"],
        model="sonnet"
    ),
    "code-auditor": AgentDefinition(
        description="Security, performance, dead code, accessibility audits",
        prompt=open("agents/code-auditor.md").read(),
        tools=["Read", "Glob", "Grep"],
        model="opus"
    ),
    "git-tracker": AgentDefinition(
        description="Git workflow — branches, PRs, merges, diffs, sync",
        prompt=open("agents/git-tracker.md").read(),
        tools=["Bash", "Read", "Grep", "Write"],
        model="sonnet"
    ),
    "knowledge-guide": AgentDefinition(
        description="Methodology guidance, note quality, connection suggestions",
        prompt=open("agents/knowledge-guide.md").read(),
        tools=["Read", "Grep"],
        model="haiku"
    ),
    "team-lead": AgentDefinition(
        description="Multi-agent team coordination, task decomposition, parallel dispatch",
        prompt=open("agents/team-lead.md").read(),
        tools=["Read", "Write", "Glob", "Grep", "Bash", "Task"],
        model="opus"
    ),
}

10x-Code — 10x-Code v2.0.0