Reflect Turns Workspace History Into Actionable Improvements

Dawn now has /reflect: a slash command that reviews recent workspace context and turns repeated friction into concrete improvement proposals. Run it when you want Dawn to look back at what has been happening, compare it with memory pages and reusable workspace assets, and suggest what should be cleaned up or made reusable.

Most teams accumulate small bits of useful context in chat. Someone explains a setup step. An agent repeats the same mistake twice. A memory page becomes stale after a project changes. A workflow is discussed several times but never captured. /reflect gives Dawn a deliberate way to inspect that recent history and report back with practical next actions.

A Workspace Audit From Chat

Type /reflect in Dawn and the agent starts a background audit of the current workspace. You can leave it broad, or add a focus in plain language:

/reflect onboarding and setup friction

/reflect check memory for stale Azure DevOps assumptions

/reflect after the billing migration

Dawn looks for recent conversations, relevant memory pages, existing skills, prompt commands, and workflows that match the focus. The result is a reflection report: what it found, what evidence supports it, what memory may need updating, and what reusable skills, commands, or workflows could help.

Evidence Before Suggestions

/reflect grounds suggestions in workspace evidence, not pattern-matched guesses.

If Dawn says a memory page may be stale, it should point to the memory page and the newer conversation that contradicts it. If it suggests a new skill, it should explain the repeated pattern it saw. If it suggests a workflow, it should connect that suggestion to actual recurring work, not just a vague idea that automation might be useful.

That makes the report easier to trust. You can see why Dawn made the suggestion and decide whether it is worth acting on.

Memory, Skills, Commands, and Workflows

/reflect is especially useful because it knows the difference between the places reusable knowledge should live.

Shared team facts, decisions, runbooks, and project conventions belong in workspace memory. Personal preferences and cross-workspace working conventions belong in personal memory. Repeatable agent instructions belong in skills. Chat shortcuts belong in prompt commands. Event-driven work belongs in workflows.

That distinction matters — each improvement proposal names the right product primitive for the job, rather than collapsing everything into “add another note somewhere.”

Proposal-Only By Default

The first version of /reflect does not change your workspace by itself. It reads accessible workspace context and returns proposals. It does not update memory pages, create skills, edit commands, start workflows, or post to connected sources during the audit run.

That keeps it safe to use when you are unsure what Dawn will find. Run /reflect, review the report, then choose which proposals to apply.

When To Use It

Use /reflect after a busy week, after a migration, after repeated support or setup questions, or whenever a workspace feels like it is carrying too much tacit knowledge in chat.

It is also useful before creating new reusable assets. Instead of guessing whether a new skill or workflow is needed, ask Dawn to inspect the recent pattern first. If the evidence is thin, the report should say so. If the pattern is clear, you have a stronger starting point.

Getting Started

Open a Dawn chat and type /reflect. Start broad if you want a general workspace improvement report, or add a short focus if you already know the area you want Dawn to inspect.

The output is a starting point, not an automatic cleanup pass. Review the suggested memory updates, skills, commands, and workflows, then apply the ones that match how your team actually wants to work.