agentleFS
aws-solutionsaws-solutions/konductor/AGENTS.md

konductor

AGENTS.md11 starsChanged 49 days agoApache-2.0View on GitHub

Guidance for custom agents on Kiro CLI or Claude Code. This file is loaded into agent context at session start. Follow these instructions for the duration of the session. Before starting work, load available project context: If a path does not exist, skip it silently and continue. Konductor is an open-source multi-agent orchestration framework that provides a coordinated suite of specialist AI agents automating the full software development lifecycle. Agents collaborate through a structured delegation protocol. The package ships as…

The file

We split this AGENTS.md into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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. Distinctive:

    Before starting work, load available project context:

    line 8
  2. Distinctive:

    If a path does not exist, skip it silently and continue.

    line 12
  3. Very specific:

    Run from the repo root (make synth's konductor synth defaults its source tree to the current working directory, and the build/cli/konductor …

    line 39
  4. Distinctive:

    License headers: All code files must carry an SPDX short-form identifier as the very first line (before any docstring or comment block), …

    line 70
  5. Distinctive:

    Use the native single-line comment syntax for the language:

    line 76
  6. Very specific:

    Use .github/PULL_REQUEST_TEMPLATE.md for every pull request.

    line 86
  7. Very specific:

    When changing the command surface, keep cli/README.md's command list in sync.

    line 101

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “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
AGENTS.md

AGENTS.md

Guidance for custom agents on Kiro CLI or Claude Code. This file is loaded into agent context at session start. Follow these instructions for the duration of the session.

Context Loading (MUST follow at session start)

Before starting work, load available project context:

  • You MUST read .konductor/memory/*.md if present: persistent workspace facts, decisions, and preferences.

If a path does not exist, skip it silently and continue.

Project Overview

Konductor is an open-source multi-agent orchestration framework that provides a coordinated suite of specialist AI agents automating the full software development lifecycle. Agents collaborate through a structured delegation protocol. The package ships as static agent configuration files compatible with Kiro CLI and Claude Code — no runtime infrastructure required.

Project Structure

agents/        # Agent definitions (.agent-spec.json, one per agent)
skills/        # Skill definitions (SKILL.md + optional scripts)
agent-sops/    # Standard operating procedures (user-invoked workflows)
context/       # Context files loaded at agent startup (system prompts, routing rules)
cli/           # Konductor CLI; see cli/README.md

Setup & Commands

# Kiro CLI
konductor install
kiro-cli chat --agent konductor

# Claude Code
claude --agent konductor

Run from the repo root (make synth's konductor synth defaults its source tree to the current working directory, and the build/cli/konductor path below is relative to it):

# Build cli/ + mcp/, synth agent/skill content, then install from this checkout
make build
make synth
build/cli/konductor install --from . --harness kiro-cli-v2

make build first is required, not optional: make synth on its own only builds cli/ (it needs the konductor binary, nothing from mcp/), so skipping this step and going straight to make synth leaves mcp/'s MCP server binary unbuilt -- install auto-discovers that binary and silently skips it if missing (no error), producing agents that can't load skills at runtime.

See cli/README.md for the full build, PATH setup, and install instructions.

Code Style & Conventions

  • Agent specs are JSON; keep them formatted (2-space indent).
  • Skills are markdown with YAML frontmatter (name, description). A description may take a trigger-clause form ("Use when...") or a behavior-summary form ("Does X"); either is fine as long as a reader, human or model, can tell when the skill applies from the text alone. Do not convert a description from one form to the other for consistency alone. Fix a description that gives no activation condition at all. You may broaden an existing condition (for example, add a disjunct) when the skill's actual scope grew, but do not churn the form otherwise.
  • Orchestrators are named konductor, konductor-mux-orchestrator, and konductor-cmux-orchestrator; every specialist uses a k-* name. Skills are unprefixed unless avoiding a known collision.
  • License headers: All code files must carry an SPDX short-form identifier as the very first line (before any docstring or comment block), matching the project's declared Apache-2.0 license. Apache-2.0 is the only permitted SPDX identifier in this package. Do not introduce any other identifier (MIT, BSD, ISC, a dual-license expression, or any other SPDX string). Exception: scripts that require a shebang (#!) line must place the SPDX header on line 2, immediately after the shebang, so the OS interpreter directive remains the literal first line. Use the native single-line comment syntax for the language:

    • Rust: // SPDX-License-Identifier: Apache-2.0
    • Python: # SPDX-License-Identifier: Apache-2.0
    • JS/TS, Go: // SPDX-License-Identifier: Apache-2.0
    • Shell, YAML: # SPDX-License-Identifier: Apache-2.0

    Formats that have no comment syntax at all (JSON and similar) cannot carry a header and are exempt -- do not add one and do not treat its absence as a violation.

Pull Requests

Use .github/PULL_REQUEST_TEMPLATE.md for every pull request. GitHub pre-fills the PR body with it automatically. Complete every section, or delete it if it does not apply — do not leave a section untouched with its placeholder text still in place.

Konductor CLI (cli/)

The cli/ tree is the Konductor CLI: the command-line utility for installing, configuring, and diagnosing a Konductor-managed repository, covering an 8-command surface. Read cli/README.md before working on it (commands, conventions, current state).

Non-negotiable conventions:

  • Usage errors exit 64 (EX_USAGE); exit code 2 is reserved for an unresolved finding at or above the effective gate threshold (CRITICAL by preset, HIGH under the strict floor for phase-gated SOPs) and must never be emitted for a bad CLI invocation.
  • When changing the command surface, keep cli/README.md's command list in sync.

Authoring Agents & Skills

See the agents-md-authoring skill for creating and maintaining AGENTS.md files.

Memory

Local memory persistence is provided by the persistent-memory skill. Facts persist across sessions in .konductor/memory/MEMORY.md (project facts) and .konductor/memory/USER.md (preferences), validated against limits and an optional URL allowlist configured in .konductor/memory-config.json.

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.