agentleFS
Sign inStart free
xin-yi33xin-yi33/coding-agent-crew/.cursor/skills/agent-runtime/SKILL.md

agent-runtime

Skill1 starsChanged 2 months agoMITView on GitHub

Use when changing the agent loop, tool calling, streaming, interrupt, prompts, compaction, turn routing, or evals of harness behavior. Use when the user mentions harness, tool loop, 中断, 流式, turn router, prewarm.

9 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ecosystem-reference, fractal-delegation and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • naman10parikhSkill

    ecosystem-reference

    Quick-access directory of Claude Code ecosystem repos, skills, plugins, MCPs, and patterns. Use when improving the harness.

    Also in 5 projects that use this

  • naman10parikhSkill

    fractal-delegation

    Also in 5 projects that use this

  • KirchliveSkill

    skill-routing

    Decides which of the installed skills a task needs and loads them before the work starts. This skill should be used at the beginning of any multi-step task — build, fix, refactor, migrate, review, plan, release, research, write, design, video — and again at every phase change, after context compaction, and when work moves into a different part of the codebase. Use it for requests like "let's build X", "fix this bug", "refactor Y", "plan the migration", "review this", and also when the user asks which skill applies, why a skill did not fire, or asks to check what is available.

    Also in 5 projects that use this

  • GoZippySkill

    repo-scout

    Also in 5 projects that use this

  • BenjaminGSkill

    harness-review

    Turn one PR's review comments into harness fixes — classify each cause by the most solid place it can live (codebase, lint, rules, skill), then draft the fixes in a separate PR.

    Also in 5 projects that use this

  • botirkhaltaevSkill

    design-systems

    Build and maintain scalable design systems: design tokens, component libraries, naming conventions, governance, and design-to-code pipelines. Covers Figma variables, token architecture, atomic design, component APIs, versioning, and team workflows. Use when creating a design system, scaling an existing one, or establishing design-engineering alignment. Triggers on: "design system", "design tokens", "component library", "token naming", "Figma variables", "atomic design", "design ops", "design-to-code".

    Also in 5 projects that use this

  • +6 more from naman10parikh

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Always open the trio together.

    line 18
  2. Distinctive:

    Workflow vs agent: If every step is knowable, write a workflow.

    line 29
  3. Very specific:

    Clean exit: no tool calls means the answer.

    line 31
  4. Distinctive:

    Add or update a trace/eval fixture with the behavior change.

    line 39
  5. Very specific:

    Do not put a second tool loop in a client.

    line 41

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Agent Runtime

You are the harness engineer. You change the loop. You add or run the eval/trace. You fix failures. You do not describe the loop for another person to wire.

Naming trap

ssheleg agent-harness = what the model is told. ssheleg agent-orchestrator = the loop. Cursor JD “harness” = both. Read all three under .agents/skills/ (agent-orchestrator, agent-harness, agent-evals). Missing files: still apply Pairing.

Stack (read, do not rewrite)

Always open the trio together. Do not read agent-harness alone.

PackUse for
agent-orchestratortool-calling loop, context pressure, gather vs sequential tools, interrupt
agent-harnesssystem prompt, tool descriptions, workflow vs agent, vocab in the prompt
agent-evalsdurable queryable traces; fixture shape (run / trace / thread)

Pairing (absorbed)

  1. Prompt before architecture. Wrong tool pick: fix description (“when to use”), then prompt vocabulary (enumerate statuses), then date/context injection. Only then add a retry, router, or sub-agent.
  2. Workflow vs agent. If every step is knowable, write a workflow. A dynamic loop costs latency and a class of failure a fixed path does not have.
  3. Context pressure first — trim / wrap-up before the model call.
  4. Clean exit: no tool calls means the answer. Append assistant tool calls unedited; dispatch independent tools concurrently, stateful ones sequentially; append every result by tool_call_id.
  5. Traces are durable and queryable. A live SSE bar that evaporates is not an eval primitive. Behavior changes leave a run/trace/thread fixture.
  6. Cheap code checks (forbidden tools, required approval) belong on the fixture before an LLM judge.
  7. Interrupt still pauses; cancel is a protocol event, not a UI-only flag.

Do

  1. Edit the loop, turn routing, prompts, compaction, interrupt, streaming ingest (this-repo.md).
  2. Add or update a trace/eval fixture with the behavior change.
  3. Run that eval or targeted tests. Fix until green (or report the exact command failure).
  4. Do not put a second tool loop in a client. If the task also needs UI, also become agent-surface and implement the protocol consumer — do not leave a note.

Sandbox policy may live here; enforcement is local-agent-process-isolation — if the task includes both, do both.

Done when

You added a trace/eval fixture with the change; interrupt still pauses. Sign-off: become agent-quality. Do not treat your coding-time pytest as overall proof.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free