These agent skills implement patterns in the Chiron Codex Patterns in AI-Augmented Software Development catalog.
- ambiguity-resolution: Identifies and resolves design conflicts and ambiguities in requirements or specifications to ensure a clear foundation before implementation. Implements Disclose Ambiguity.
- analyze-interaction-transcript: Diagnoses systemic failures in the collaboration process by analyzing interaction dynamics. Implements You Reflect and Model Reflects.
- conduct-bake-off: Replaces subjective "vibes" with a controlled experiment to objectively compare two alternatives. Implements Bake Off.
- create-execution-plan: Transforms high-level goals into detailed, risk-aware execution plans for user review before any coding begins. Implements Stop and Plan.
- design-verification-strategy: Establishes a concrete, external system of verification and validation to define a verifiable "Definition of Done." Implements Trust the Tests.
- draft-engineering-log: Synthesizes the day's interactions into a structured engineering log from the developer's perspective. Implements Remember What We Did.
- help-me-reflect: Identifies points where human emotional responses have negatively impacted interaction quality and suggests improvements. Implements You Reflect.
- identify-extractable-prompts: Decomposes complex goals into subtasks and proposes specific prompts or skills to provide more control over execution. Implements Extract Prompt.
- propose-deterministic-tool: Identifies the "deterministic core" of a problem to replace unreliable generative processes with reproducible tools. Implements Replace Vibes with Tools.
- side-by-side: Supports scientific exploration of human-AI interactions by delegating the same task to multiple parallel subagents for comparison. Implements Bake Off.
- take-baby-steps: Decomposes large, ambiguous goals into a sequence of small, verifiable subtasks to minimize divergence and maximize feedback speed. Implements Baby Steps.
- task-review: Compares the current state of a project against initial instructions to determine if the task is complete or if further work is required. Implements Model Reviews.
- triage-session-findings: Distills long conversation histories into actionable assets, reusable prompts, and technical memories. Implements Extract Prompt, Write Down TODOs, and Remember What We Did.
- write-manual: Generates clear, comprehensive, and unambiguous user instructions for software deployment and operational flows. Implements Generate Documentation.
Clone this repo, and copy the skills definitions into your agent's skills folder (for example, ~/.agent/skills/), or your project's agent skills folder (for example, /path/to/project/.agent/skills).
Support for agent marketplaces will come soon!
Chiron Codex makes these skills available under the terms of the GNU Affero General Public License version 3.0, or any later version accepted by the project's proxy. Pursuant to section 14 of the license, Dr. Graham Lee is the nominated proxy for accepting later versions.
If you require access to the skills under alternative license arrangements, discuss these with Dr. Graham Lee: use the email address in the commit messages, or send a direct message in the Chiron Codex Patreon.
For information about these agent skills, or to help their continued development, join the Chiron Codex Patreon.