AGENT SKILL

opensquilla

Route OpenSquilla installation, gateway, provider and SquillaRouter configuration, CLI automation, channels and MCP, Skills and MetaSkills, and TUI or desktop work to the right bundled sub-skill.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
opensquilla
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Route OpenSquilla installation, gateway, provider and SquillaRouter configuration, CLI automation, channels and MCP, Skills and MetaSkills, and TUI or desktop work to the right bundled sub-skill.
disable-model-invocation
true

02 资源文件 7

03 SKILL.md 正文

OpenSquilla Repo Skill

Use this skill when a task names OpenSquilla or its opensquilla CLI, gateway, SquillaRouter, provider/search configuration, sessions and automation, messaging channels, MCP bridge, Skill/MetaSkill catalog, terminal UI, Web UI, or desktop runtime. This root is a router: load the narrowest sub-skill before prescribing commands or changing state.

Route Map

User intent Read
Install OpenSquilla, run onboarding, start or diagnose the gateway, open the Web UI, or choose release versus source installation sub-skills/setup-and-gateway/SKILL.md
Configure providers, models, API-key environment variables, SquillaRouter modes, web search, or configuration precedence sub-skills/configuration-and-routing/SKILL.md
Run or automate chat/agent/code tasks, sessions, memory, cron, cost, diagnostics, replay, migration, recovery, sandbox, reset, init, bundle, dist, or uninstall sub-skills/cli-and-automation/SKILL.md
Configure messaging channels, pairings, admission, certification, delivery diagnostics, or the stdio MCP server bridge sub-skills/channels-and-integrations/SKILL.md
Discover, install, update, publish, inspect, or troubleshoot Skills and MetaSkills, taps, runs, or proposals sub-skills/skills-and-meta/SKILL.md
Diagnose terminal rendering, OpenTUI companion behavior, Web UI presentation/artifact previews, or desktop packaging and runtime sub-skills/tui-and-desktop/SKILL.md

When a request crosses boundaries, start with setup/gateway readiness, then load each relevant specialist. Keep UI presentation separate from provider routing, and keep channel transport separate from ordinary chat/session automation.

Safe First Checks

Prefer read-only inspection before state changes:

opensquilla --help
opensquilla onboard status
opensquilla gateway status
opensquilla doctor

The bundled scripts/runtime-health.sh performs a safe local snapshot. It never starts a gateway, changes configuration, calls an LLM/search provider, installs dependencies, or mutates the Skill catalog.

Root References

Operating Rules

  1. Treat provider keys, channel tokens, webhook secrets, exported sessions, raw diagnostics, and content-bearing bundles as sensitive. Prefer environment variable references and redacted output.
  2. Distinguish local catalog/schema checks from live readiness. A listed provider, search adapter, or channel type is not proof that credentials or network access work.
  3. Do not expose a gateway beyond loopback without an explicit operator request, authentication, and a reviewed reverse-proxy or tunnel boundary.
  4. Call out state-changing commands before use: configuration edits, gateway restart/reload, Skill install/update/uninstall/reload/publish, proposal accept, cron mutation, migration apply, reset, and uninstall purge options.
  5. Use release-install guidance for ordinary users. Source-build, browser, Electron, Bun/Node, and maintainer harness instructions apply only to a checkout.
  6. This bundle targets OpenSquilla 0.5.3. Re-check current command help and the provenance reference when operating a materially different release.

Verification Boundary

The skill was built from source inspection and CPU-only installed-package checks. CLI help, selected imports, and local catalogs were checked; no credentialed LLM, search, channel, external MCP client, browser/Electron end-to-end, or public gateway test is implied. Never turn those live checks on merely to validate this skill bundle.

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351