AutoSaddler is a system that improves LLM-agent harnesses by diagnosing execution traces and applying structured changes to prompts, tools, middleware, and agent-loop logic. It evaluates candidate updates for their ability to generalize beyond the traces that motivated them. The catalogue add-ons represent workflows for using AutoSaddler.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add microsoft/AutoSaddler --skill steering-patchgit clone --depth 1 https://github.com/microsoft/AutoSaddlerWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/microsoft/autosaddler/steering-patch)<a href="https://agentmods.dev/skills/microsoft/autosaddler/steering-patch"><img src="https://agentmods.dev/badge/skills/microsoft/autosaddler/steering-patch/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/microsoft/autosaddler/steering-patch"><img src="https://agentmods.dev/badge/skills/microsoft/autosaddler/steering-patch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
SkillSpector: 1 finding, up to low
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- low Excessive Agency · line 58 Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.Fix: Limit the skill's scope to its documented purpose. Remove instructions that enable the agent to perform actions outside its stated functionality.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00029 | $0.01430 |
| Opus 5.5 | $0.00012 | $0.00572 |
| Sonnet 5.5 | $0.00006 | $0.00286 |
| Haiku 4.5 | $0.00003 | $0.00143 |
Grade A, and why
steering-patch scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 158 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Steering Patch
Overview
A steering patch refines the agent's behavior within the existing capabilities. It modifies how the agent uses existing tools — through prompt rules, tool description corrections, or hook reminders.
Key distinction from capability patches: A steering patch does not add
new tool methods, new parameters, or modify tool implementation code.
It only changes the text the agent reads (docstrings, prompts, hooks).
The outer loop selects which skill to use based on the current phase:
steering phase uses this skill, capability phase uses capability-patch.
When to Use
Based on the diagnosis results, generate a patch that resolves the identified root cause. Typical root causes addressable by steering patches:
| Root Cause Category | Signal | Example |
|---|---|---|
| Misleading description | Docstring doesn't match implementation | Agent uses correct tool but passes wrong parameter format |
| Missing description detail | Docstring lacks constraints or semantics | Agent has correct intent but doesn't know about a required field |
| Behavioral pattern failure | Multiple scenarios share the same mistake | Agent consistently forgets to check preconditions before acting |
| Missing runtime guidance | Agent needs just-in-time hints for specific tools | Agent misuses a tool because it doesn't recall a constraint at call time |
Workflow
Step 1: Consult History
Before writing any changes, check what has been tried before:
# Full patch history: diffs, reflections, lessons (good/bad patterns)
evo-dag show history
# Per-scenario history: prior root causes and attempted fixes
evo-dag show scenario <scenario_id>
- Build on proven strategies: Which steering patches generalized well to the dev set? What root-cause patterns did they resolve?
- Avoid known bad patterns: Which patches caused regressions or failed to generalize? Which rule phrasings were too weak or too aggressive?
- Do not re-attempt failed approaches: If a specific fix was already tried on this scenario and failed, try a different approach.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 158 lines · 29 tokens per session scan A 1cf9a9ef54e3
steering-patch is a skill published in the GitHub repository microsoft/AutoSaddler (243 stars, last pushed 3d ago), licensed MIT. It adds 29 tokens to every session and 1,430 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
local-ai-agents
Build local-first AI agents that run entirely on a developer workstation with Microsoft Foundry Local and Qwen function-calling models. Covers Small Language Models (SLMs), the OpenAI-compatible local endpoint, sandboxed local tools, local RAG with Chroma, local MCP servers, hybrid cloud/local routing, and the…
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
next-cache-components-adoption
Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…
next-cache-components-optimizer
Optimize the meaningful UI available immediately from a Next.js route on an initial load (hard navigation) or named client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route and entry point at a time; the shipped test then guards…