agentsclimarketplace

Vaultspec adr

Skill nevenincs/vaultspec-core/src/vaultspec_core/builtins/skills/vaultspec-adr

A spec-driven harness for coding agents (and, humans)

Install
npx -y skills add nevenincs/vaultspec-core --skill vaultspec-adr

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Capture an architectural decision as an ADR in .vault/adr/. Use after research, before planning, when a significant design choice and its trade-offs must be recorded.

SKILL.md

5.0 KB, as published. Nobody here has run it

ADR writing skill (vaultspec-adr)

Use this skill:

  • After a vaultspec-research session has concluded with a recommendation.
  • When multiple competing technical choices need a grounding document.
  • When a significant architectural decision is made that affects the project's foundations, feature set, or development trajectory.
  • To document the blast radius, "why", and "what" of major architectural choices.
  • When an accepted decision needs refining or concretizing - this amends the existing record in place, not a new document (see Amend or supersede).

Announce at start: "I'm using the vaultspec-adr skill to create a new ADR."

Required steps

  • Ground the decision in existing intent first. Before drafting, retrieve the prior decisions that bind this area - the surest way to architect on top of them rather than silently contradicting them. Lead with vaultspec-rag search "<intent>" --type vault --doc-type adr (the directed ADR filter, sharper than catch-all --type vault); read each in full - a record already governing this scope routes through Amend or supersede below. Use vaultspec-rag search "<intent>" --type code for the implementation sites the decision will touch, then confirm exact symbols with grep, and lean on vaultspec-core status and vaultspec-core vault list - first-class for orientation - to map related records. Round out decision recall by listing .vault/adr/ and filtering by feature - search can miss lower-ranked or opaquely-named records. Where vaultspec-rag is not installed, the vaultspec-core discovery verbs and grep carry the same sequence.

  • Read and use the template at .vaultspec/templates/adr.md; its embedded hint blocks govern the body structure.

  • Scaffold via the CLI: vaultspec-core vault add adr --feature {feature} --related <research-stem>, then author the body prose in the scaffolded file. The CLI owns the filename (.vault/adr/yyyy-mm-dd-{feature}-adr.md) and the frontmatter; never hand-write either. The full frontmatter schema is defined in the vaultspec rule; verify after scaffolding with vaultspec-core vault check all rather than hand-editing frontmatter.

  • Read and link related Research from: .vault/research/yyyy-mm-dd-{feature}-research.md.

  • Terminate if related research is not found and prompt the user to first invoke vaultspec-research.

Amend or supersede

One decision, one governing record. The grounding step surfaces the ADRs that already govern this scope; when one exists, choose the mode before touching anything:

  • Amend (the default). Refinement, concretization, narrowed scope, an implementation-driven parameter change: rewrite the existing record's body in place, with the same user approval a new ADR would need. Status stays accepted; the modified: stamp carries the revision. No new document.
  • Supersede (pivot only). The decision reverses direction, or the record's rationale no longer supports the new choice: scaffold the new ADR and in the same session run vaultspec-core vault adr supersede OLD --by NEW, so exactly one record is accepted for the scope.

Refinement-by-new-document is the failure mode this rule exists to stop: it piles sibling ADRs onto one decision until the cluster carries mutually contradictory accepted markers and the supersession chain, not the decision, becomes the thing being managed.

Document boundary

Each fact has one home: the research grounds, the ADR decides. The record cites research and reference findings by stem (e.g. 2026-02-04-editor-demo-research) and never restates their evidence - a restated fact forks, bloats agent context, and goes stale silently. If drafting surfaces a fact the grounding lacks, add it to the research or <Reference> first, then cite it. Conversely, the decision lives only here: strip any decision language the grounding documents carry into this record rather than duplicating it.

Workflow

  • Derive from Research: ADRs should always be preceded by a vaultspec-research session.

  • CRITICAL: you MUST always present ADR findings as an interactive prompt for user approval. Without explicit user sign-off the ADR is considered null and void.

  • Draft the ADR using the vaultspec-adr-researcher agent persona, which formalizes the research-backed decisions into ADR content and returns it for persistence into the scaffolded document (the vaultspec-writer persona's mandate is plan-only).

  • Associate the ADR with {feature} based on the findings in [[...-research.md]].

  • Supersession mechanics: vaultspec-core vault adr supersede OLD --by NEW records the superseded_by: back-pointer and rewrites the old record's status; never edit status lines by hand.

  • Linking: Use [[wiki-links]] for references. DO NOT use @ref or [label](path).

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.