Askit Migrate

askit-migrate

Purpose

Adopt an existing skills repository into a Standard-conformant plugin. Three modes: assess surveys the repo's components and structure, maps them to the Standard's component types, and reports the conformance gap (what is missing for Bronze, then Silver), handling the pre-library.json state the conformance core cannot yet grade; plan produces an ordered bring-to-conformance roadmap keyed to the tiers and names the askit-build-* skill or check that closes each gap; adopt performs the first structural step by writing a minimal canonical library.json and a root AGENTS.md if absent, so the repo becomes gradeable and the rest of the plan can run through askit-evaluate and the builders. Authoring depth is in [references/migration-workflow.md](references/migration-workflow.md).

When to use

When adopting a foreign or ad-hoc skills repo, migrating a Claude-only plugin toward cross-agent conformance, or planning a Bronze-to-Silver upgrade path for an existing repository.

assess mode

1. Discover what exists: skills, commands, subagents, hooks, MCP config, instructions (AGENTS.md / CLAUDE.md), and any manifest. Map each to a Standard component type (sec 3).
2. Report the gap by tier. The Bronze blockers first (a parseable library.json, a root AGENTS.md, valid skill frontmatter, name-equals-directory, resolvable reference links), then the Silver blockers (agent-targets, the plugin prefix, a components index, native manifests).
3. Where a library.json already exists, delegate the per-rule grading to askit-evaluate; otherwise report the pre-manifest gaps the conformance core cannot yet see and run assess again after adopt.

plan mode

1. Order the gaps into a staged roadmap: Bronze first (make it parse and self-describe), then Silver (cross-agent shape and the prefix), then the optional tail (hooks, workflows, docs site).
2. For each step name the resolver: the askit-build-* skill that authors the component, the check whose message states the fix, or askit-evaluate to re-confirm.
3. Record the plan so the migration is legible and resumable across sessions.

adopt mode

1. Write the Bronze-minimal canonical library.json the conformance core requires at every tier (sec 5.1, check U1): name, version 0.1.0, description, standard (the Standard version targeted), and tier universal. That is the smallest manifest that satisfies U1 and makes the repo gradeable.
2. Add a root AGENTS.md if absent (Standard sec 3.10) so anatomy (U2) passes, then run askit-evaluate to confirm Bronze is clean.
3. When climbing to Silver, add the Convergent-only fields the higher checks require: agent-targets (S1), a kebab-case prefix ending in a hyphen carried by every component name (S2), and a components index (S3); then emit native manifests with node scripts/generators/gen-manifest.mjs . --write --target=all (S6) and hand off to the askit-build-* skills and askit-evaluate.

Scope

askit-migrate adopts an existing repo; greenfield scaffolding from an interview is a separate, planned init/onboarding flow (Standard sec 10.7). It makes the repo gradeable and produces the plan, then the askit-build-* skills and askit-evaluate execute the plan slice by slice; it does not rename components to add the prefix wholesale or rewrite skills on its own. The why-gate applies before adding any new component (see askit-backlog).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: product-on-purpose
• Source: product-on-purpose/agent-skills-toolkit
• License: Apache-2.0
• Homepage: https://product-on-purpose.github.io/agent-skills-toolkit

Install and usage instructions live in the source repository linked above.