Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.driftless.icu/llms.txt

Use this file to discover all available pages before exploring further.

Driftless installs repo-local instructions for coding agents. The instructions teach agents to pull team context before editing, inspect impact, and persist durable discoveries.

Install

driftless install-skill
driftless init runs this automatically for new repos.

Files written

FilePurpose
.driftless/skill.mdDogfood copy of the Driftless agent workflow
AGENTS.mdAgent instructions used by Codex and other agentic tools
CLAUDE.mdAgent instructions used by Claude Code
The managed block is wrapped in comments:
<!-- driftless:start -->
...
<!-- driftless:end -->
Re-run driftless install-skill to refresh only the managed Driftless block.

Expected agent workflow

1

Sync on session start

driftless sync
The agent sees stale topics, changed context, PR activity, and suggested topics before touching files.
2

Load context before editing

driftless context load --files "src/auth/guard.ts,src/users/service.ts"
The agent gets matching topics, coverage, and recommended actions for the files it is about to modify.
3

Check blast radius

driftless graph impact --files "src/auth/guard.ts,src/users/service.ts"
The graph shows upstream and downstream components affected by the change.
4

Persist discoveries

driftless context update auth-flow \
  --decision "WorkspaceGuard is global; use @Public() only for explicit public routes" \
  --gotcha "Do not assume routes are public when guards are composed"
Durable context goes back to Cloud so the next session does not start from zero.

What to write back

Use Driftless for durable architecture knowledge:
  • Decisions that explain why the code works this way
  • Gotchas that would cause a future agent to make a bad edit
  • Invariants that must stay true across refactors
  • New or corrected anchors after code moves
  • Cross-repo relationships discovered while working
Do not store secrets, credentials, raw customer data, or temporary scratch notes in context topics.