Skip to content

Skills overview

agent-os ships its skills in three buckets. Workflows are invoked by hand, disciplines trigger themselves from the situation, and the meta-skill governs how the others are written. Read this page top to bottom: the planning foundation, the on-ramp in front of it, the remaining workflows, and finally the component skills the workflows compose.

The foundation: shape and chart

Planning starts here, and each layer builds on the one before it. shape-work turns one bounded idea into implementation-ready issues. chart-work builds on it: a broad effort becomes decision tickets, and every bounded branch ends exactly where shape-work begins.

SkillBucketInvocationPurpose
shape-workworkflowmanualTurn bounded choices into implementation-ready issues
chart-workworkflowmanualChart broad work as a parallel graph of decision tickets

The on-ramp: guide-me

guide-me builds on both. When you cannot state the goal yet, it only shows the way: questioning through its component understand-work, the plain-language gate through explain-work, and — once you approve the summary — it ends where chart-work or shape-work begins.

SkillBucketInvocationPurpose
guide-meworkflowmanualGuide a vague desire to an approved goal and into planning

Execution and operations

The workflows that run, pick, set up, and remember work.

SkillBucketInvocationPurpose
deliver-workworkflowmanualImplement one change against boundaries and ground truth
batch-workworkflowmanualRun isolated ready units and verify the integrated result
dispatch-nextworkflowmanualPick or dispatch one action according to the request
init-agent-osworkflowmanualManaged policy setup or repository defaults
record-lessonworkflowmanualRecord a durable lesson in repo or global policy
simplifierworkflowmanualRemove unnecessary code and solution layers
simplifier-auditworkflowmanualAudit a repository for simplification opportunities

Component skills

Standalone pieces that guide-me composes. Invoke them directly whenever the piece is useful on its own — a grilling without the routing, or a plain-language summary of any plan, diff, or pull request.

SkillBucketInvocationPurpose
understand-workworkflowmanualQuestion out the need behind a stated wish
explain-workworkflowmanualExplain the task in plain language for approval

Disciplines

Always on; no invocation.

SkillBucketInvocationPurpose
verify-before-donedisciplineautomaticFresh evidence before any completion claim
diagnose-before-fixdisciplineautomaticReproduce and root-cause before patching
proportional-testingdisciplineautomaticMinimum meaningful regression coverage
scope-guarddisciplineautomaticKeep work inside the task; flag drift
simplifier-reviewdisciplineautomaticReview a diff for unnecessary complexity
notice-lessondisciplineautomaticTreat interruptions as misunderstanding signals

Meta

SkillBucketInvocationPurpose
list-skillsmetamanualList installed skills and how to invoke them
writing-skillsmetamanualDoctrine and definition of done for agent-os skills

Invocation

Manual skills are opted into explicitly on both platforms. In Claude Code they carry disable-model-invocation: true in their frontmatter and are typed as /<skill> when installed directly or /agent-os:<skill> through the plugin. In Codex they carry an agents/openai.yaml with policy.allow_implicit_invocation: false and are typed as $<skill>.

Automatic skills carry neither switch. Their descriptions name the situation and boundary clearly enough to trigger only when useful.

Retirement

A retired skill moves to the repository's root deprecated/ folder rather than being deleted. That folder sits outside the plugin's skills/ directory, so its contents are never distributed, and a skill re-enters skills/ only by passing the definition of done again.

A personal framework, published in the open.