ad-state-migrate
Skill0 starsChanged 6 days agoMITView on GitHub
Use when the user wants to move Agent Daemon (ad) to another computer or between the OSes of a dual boot — \"move ad to my new laptop\", \"export my ad memory\", \"import ad state\", \"Windows se Linux pe ad le jao\", \"naye computer pe ad setup\", \"ad ka data transfer karo\". Runs ad state export on the side being left and ad state import on the other, proves recall, explains review, re-sign-in, ad init and the dual-boot checkout rule. Never copies ~/.agent-daemon by hand.
In this repository174 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ad-ci-readinessSkill
- ad-pre-pushSkill
- accessibilitySkill
- agent-evalSkill
- agent-harness-constructionSkill
- agentic-engineeringSkill
- agent-introspection-debuggingSkill
- agent-sortSkill
- ai-first-engineeringSkill
- ai-regression-testingSkill
- api-connector-builderSkill
- architecture-decision-recordsSkill
- article-writingSkill
- automation-audit-opsSkill
- autonomous-agent-harnessSkill
- autonomous-loopsSkill
- backend-patternsSkill
- benchmarkSkill
- blueprintSkill
- brand-voiceSkill
- browser-qaSkill
- bun-runtimeSkill
- canary-watchSkill
- cavemanSkill
- ckSkill
- claude-devfleetSkill
- click-path-auditSkill
- codebase-onboardingSkill
- code-tourSkill
- coding-standardsSkill
- content-engineSkill
- content-hash-cache-patternSkill
- context-budgetSkill
- continuous-agent-loopSkill
- continuous-learningSkill
- continuous-learning-v2Skill
- cost-aware-llm-pipelineSkill
- councilSkill
- crosspostSkill
- ad-harnessSkill
- ad-tui-devSkill
- agent-daemon-memorySkill
- big-feature-flowSkill
- bootstrap-daemonSkill
- bug-flowSkill
- engine-upgradeSkill
- feature-flowSkill
- gepa-evolve-inlineSkill
- harness-troubleshootSkill
- orchestrate-teamSkill
- release-flowSkill
- session-closeSkill
- skill-authorSkill
- skill-installerSkill
- dashboard-builderSkill
- database-migrationsSkill
- deep-researchSkill
- deployment-patternsSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 2 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Never extract or edit a bundle by hand, and never copy ~/.agent-daemon or the engine's folder directly.
- Very specific: line 20
Never read secrets/, auth.json or another tool's credential store, and never ask for a key in chat.
- Distinctive: line 21
Never delete or change the state on the side being left.
- Very specific: line 22
Never claim the move worked without ad state verify --import output.
- Very specific: line 23
Never use --on-conflict take-theirs, --force-tracked or --accept-held without the user's explicit yes for this import.
- Very specific: line 24
Never run ad review --import, --accept-all or ad memory promote yourself: the user does, in their own terminal.
- Distinctive: line 25
Never put the bundle in a synced or shared cloud folder, or in git.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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.
Files
Move ad to another computer
This is not ad migrate (that only renames ad's engine folder on one machine). Moving ad means: export on the side being left, carry one file privately, import on the other side, prove recall, then the user reviews, signs in and re-runs ad init. Quote ad's output; never paraphrase a pass.
Never
- Never extract or edit a bundle by hand, and never copy
~/.agent-daemonor the engine's folder directly. - Never read
secrets/,auth.jsonor another tool's credential store, and never ask for a key in chat. - Never delete or change the state on the side being left.
- Never claim the move worked without
ad state verify --importoutput. - Never use
--on-conflict take-theirs,--force-trackedor--accept-heldwithout the user's explicit yes for this import. - Never run
ad review --import,--accept-allorad memory promoteyourself: the user does, in their own terminal. - Never put the bundle in a synced or shared cloud folder, or in git.
0. Which side, and is this a repeat move?
Ask once: "Are we on the side you're leaving (export), or the side you're moving to (import)? Have you moved between these two before (for example, the two OSes of a dual boot)?" Check ad --version on both sides and update the older one first. If you run inside ad's own tool shell, ad state refuses ("… doesn't run them from an agent's shell"): ask the user to run each command in their own terminal and paste back the output.
Side being left: export
ad state inspect --all-projects→ show the projects and classes. Ask which projects to take, and whether to add--include global-memory(included with--all-projects),--include engine-rollouts(large, view-only),--include tui-historyor--include project-journals.ad state export --out <path>.adstate --project <dir>(one--projectper project, or--all-projects). If it stops on items that look like keys, show the list it printed and let the user choose:--redact(cleans the export only) orad memory redact <table>:<id>(cleans their own copy, in their terminal).ad state verify <path>.adstate→ must say "no problem found". On a problem: stop, show it, and retry at most once with different flags.- Carry the file: it is not encrypted (the export says so). Use the user's own disk: for a dual boot, a partition both OSes can read; otherwise a USB drive.
Side being moved to: import
- Install ad (
docs/installation-guide.md). Dual boot: one checkout per OS (or a global install per OS): a shared checkout's native SQLite module, rebuilt for one OS, breaks the other. Thenad doctorshows the SQLite driver working. - Clone the projects with git if they aren't there yet: the bundle carries ad's state, not source code.
ad state import <path>.adstate --dry-run --search <dir>→ read the plan with the user: projects not mapped (--map <name>=<dir>), files held back, conflicts. First move: keep the default (--on-conflict refuse); the user decides each conflict. Repeat move:--on-conflict keep-both(memory merges and dedups; a differing file arrives as<name>.imported-<date>beside theirs).ad state import <path>.adstate --map <name>=<dir> --on-conflict keep-both(the flags the plan needed) → note the import id it prints.ad state verify --import <importId>→ "Recall: N/N lessons found in<project>" for each project. Fewer than N: report it plainly, offerad state rollback <importId>, and don't call the move done.- Review is the user's step: imported lessons and facts are searchable now but reach sessions only after they accept them. Tell them to run
ad review --import <importId>(or--accept-all, typing the import id) in their own terminal; a fact meant for every project still needsad memory promote. - Sign in again:
ad auth list; for each account that needs it, the user runsad auth login <provider> --label <label>(a browser or a hidden prompt). ad initin each mapped project (hooks and managed blocks; engine folders rebuild on first start).- Schedules come in switched off:
ad schedule list, thenad schedule enable <id>for the ones wanted;ad service installif jobs should run here. Skills: show the reinstall list from the import with its sources (some are git URLs) and install only what the user confirms. - Undo if needed:
ad state rollback <importId>removes only what the import added and keeps anything changed since; the side left behind was never changed.
Report
Versions on both sides, projects mapped (and any left unmapped), what came in, the recall result word for word, items waiting for review or held back, accounts still to sign in, and the manual steps left (review, ad init, schedules).
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
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.

