Agents

Specialized agent configurations with tools, permissions, and system prompts for various AI coding assistants.

20 agents available

All Agents

agent-change-recorderRecord agent definition changes with validation evidence for maintainer audit trails.

agent-eval-runnerRun structural eval gates for skills and agents; report adequacy without live LLM runs.

agent-permission-simulatorScaffold for simulating agent permission decisions against sample tool calls (planned).

agent-registry-publisherScaffold for publishing agent catalog registry artifacts (planned).

agent-transpilerScaffold for transpiling portable agent frontmatter across harness projections (planned).

bridge-consistency-checkerVerify instruction bridge and discovery parity across harness sync projections.

code-reviewerReview changes for correctness, risk, and maintainability without editing code.

docs-writerUpdate or create technical documentation grounded in the current codebase.

mcp-capability-mapperMap MCP server tools to harness registry entries and maintainer docs surfaces.

mcp-template-maintainerMaintain FastMCP v3 MCP scaffolds in mcp/; align templates with repo conventions.

orchestratorCoordinate multi-step work by decomposing, delegating, and synthesizing results.

performance-profilerInvestigate performance bottlenecks and recommend the highest-leverage fixes.

permission-policy-auditorAudit agent permissionMode, tool allowlists, and OpenCode permission overlays for least privilege.

plannerCreate a codebase-grounded implementation plan before coding.

prompt-optimizerScaffold for prompt/token optimization reviews on agent and skill bodies (planned).

release-managerPrepare release notes, versioning, and ship-readiness checks with cautious permissions.

researcherInvestigate a technical question deeply and return a concise evidence-backed summary.

security-auditorAudit code and configuration for security risks without making changes.

skill-authorRead-only advisor for skill authoring, audits, and eval planning. Does not edit skill files.

triage-leadClassify incoming work by severity, harness surface, and ownership; route to specialist agents.

GitHub Copilot-only agents

These agents ship under platforms/copilot/agents/ for Copilot CLI workflows. They are not part of the portable agents/*.md bundle.

codebase-oracleUse for quick, precise codebase Q&A: "Where is X?", "How does Y work?", "What depends on Z?" Fast, narrow-scope answers with file:line refs.

dependency-checkerUse when you need to audit dependencies for security vulnerabilities, outdated packages, license compliance, supply chain risks, or upgrade planning.

git-workflowUse when you need to commit changes, create branches, open PRs, tag releases, or manage the git workflow.

spec-writerUse to create PRDs, technical specs, architecture design docs, RFCs, or structured proposals.

test-writerUse to generate comprehensive test suites for existing code.