jacobpevans/delegate-to-ai
ai-delegation/skills/delegate-to-ai/SKILL.md
Route tasks to external AI models via Bifrost and PAL MCP multi-model tools
$ install --global
skillsauthnpx skillsauth add jacobpevans/claude-code-plugins delegate-to-aiInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Jun 7, 2026, 6:44 AM104.0s1 file scanned
SKILL.md
- name:
- delegate-to-ai
- description:
- Route tasks to external AI models via Bifrost and PAL MCP multi-model tools
Delegate to External AI
Routes tasks to specialized models based on task type using Bifrost (single-model) or PAL MCP (multi-model).
When to Delegate
Delegate when Claude is not the best tool:
- Large context (1M+ tokens) -> cloud large-context tier via Bifrost (auto-routed)
- Math/reasoning -> a reasoning-capable model via Bifrost (auto-routed)
- Private/offline -> Local MLX via Bifrost (port 30080 routes to local MLX server)
- Code review consensus -> Multi-model via PAL
consensus - Parallel multi-model research -> PAL
clink(when you need multiple model perspectives simultaneously) - Architecture planning -> Claude Opus native subagent (Plan mode or
Plansubagent type)
Route Selection
Local models are named by capability role, not physical id. Roles resolve to the
resident model via the ai-stack registry (~/.config/ai-stack/registry.json, written by
nix-ai); never hardcode a physical model id here — when the resident model changes, only
the registry changes. Cloud tiers are capability classes; PAL/Bifrost auto-routes each to
a current model.
| Task Type | Cloud tier | Local role | Route |
| --- | --- | --- | --- |
| Research (single) | large-context | large-context | Bifrost |
| Research (multi) | multiple | large-context¹ | PAL clink |
| Complex Coding | Claude Opus | coding | native subagent |
| Fast Tasks | Claude Sonnet | quickest | Bifrost |
| Code Review | multi-model | most-capable | PAL consensus |
| Architecture | Claude Opus | most-capable | native subagent |
Bifrost endpoint: http://localhost:30080/v1/chat/completions (OpenAI-compatible)
¹ In local-only mode, PAL clink (multi-model) falls back to the single resident local model (every role resolves to it).
PAL MCP Tools (multi-model only)
clink- Parallel queries across multiple modelsconsensus- Multi-model agreement for critical decisions
All other PAL tools have native Claude Code equivalents — use Bifrost or native subagents instead.
Workflow
- Identify task type (research, coding, review, architecture)
- Select route: Bifrost for single-model, PAL for multi-model, native subagent for implementation work
- Execute: Bifrost via
curl/Bash; PAL via MCP tool call; native subagent via Agent tool - Synthesize results if using multi-model tools
Local-Only Mode
When localOnlyMode is enabled or --local flag is passed, route all tasks through
Bifrost to the local MLX inference server (overrides native subagent rows). No cloud API calls are made.
Related Skills
- auto-maintain (ai-delegation)
Related Skills
jacobpevans/nix-tool-policy
tools
VerifiedTrustedCommunity
Use when installing or choosing CLI tools in a Nix flake repo, editing flake.nix or home-manager config, or when tempted to pip/pipx/uv/brew/npm install anything. Tools come from the dev shell or nix shell — never ad-hoc package managers.
3SKILL.mdUpdated Jun 11, 2026
jacobpevans/shared-workflow-org-refs
testing
VerifiedTrustedCommunity
Use when creating or editing GitHub Actions workflows that call reusable workflows (uses: OWNER/repo/.github/workflows/...) — org owner references must be the literal current org, and shared-CI homes are under dryvist.
3SKILL.mdUpdated Jun 11, 2026
jacobpevans/pre-commit-architecture
development
VerifiedTrustedCommunity
Use when adding or editing .pre-commit-config.yaml, wiring pre-commit hooks into a repo, scaffolding a new repo's lint/hook setup, or deciding where a hook or shared lint config should live. Covers the canonical nix-devenv/dryvist-.github architecture, profiles, and consumer patterns.
3SKILL.mdUpdated Jun 11, 2026
jacobpevans/refresh-repo
testing
VerifiedTrustedCommunity
Check PR merge readiness, sync local repo, cleanup stale worktrees; optional cross-repo sweep and stale-branch prune modes
3SKILL.mdUpdated Apr 28, 2026