メニュー
List inbound References and Supersedes citations of a target ADR via `adr-fmt --refs <ADR_ID>`. Use when the user asks who references / supersedes / cites an ADR, wants to assess blast radius before changing an ADR, or mentions "adr-fmt --refs".
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
| name | adr-refs |
| description | List inbound References and Supersedes citations of a target ADR via `adr-fmt --refs <ADR_ID>`. Use when the user asks who references / supersedes / cites an ADR, wants to assess blast radius before changing an ADR, or mentions "adr-fmt --refs". |
Wrapper around adr-fmt --refs <ADR_ID>. Answers "who points at this
ADR?" — the inverse of reading an ADR's own References block.
The ADR ID is a domain prefix + zero-padded number, e.g. AFM-0001,
SEC-0006, CHE-0035. Case-sensitive. If the user supplies a bare
number or filename, resolve to an ID first (read the file's front
matter or use adr-fmt --tree <prefix> to list).
cargo run -p adr-fmt --release -- --refs <ADR_ID>
Output is one record per inbound citation:
- <ADR-ID> [<Relation>] | Tier: <T> | Status: <S> | <Title>
Where <Relation> is References or Supersedes (AFM-0009 vocabulary
is exactly three verbs; Parent is implicit, not surfaced by --refs).
Target: <ADR-ID> <Title>
Inbound: References=<n> Supersedes=<n>
Citations:
- <ADR-ID> [<Rel>] T=<tier> S=<status> <title>
- ...
If the citation list is long (> ~20), summarise by tier:
By tier: S=<n> A=<n> B=<n> C=<n> D=<n>
Top (by tier strength):
- ...
| User intent | What to do |
|---|---|
| "Can I edit AFM-0009 safely?" | Run --refs; surface inbound count + tier mix |
| "Is anyone superseding X?" | Filter relations to [Supersedes] |
| "Find dead references" | Compare --refs vs file existence (out of scope for this skill — escalate) |
| "Who references my ADR?" | Use AFM-0024 pattern: --refs to find downstream impact |
Inbound: 0 cleanly.Use for any question about a codebase, its architecture, file relationships, or project content — especially when graphify-out/ exists, where the question should be treated as a graphify query first. Turns any input (code, docs, papers, images, videos) into a persistent knowledge graph with god nodes, community detection, and query/path/explain tools.
Surface the architectural rules applicable to a specific crate via `adr-fmt --context <CRATE>`. Use when an agent (especially oracle, moltke, hopper) needs the binding constraints for code in a crate before deciding or editing, or when the user asks "what rules apply to crate X?"
Run `adr-fmt --lint` against the ADR corpus and triage diagnostics by namespace (T0xx template, L0xx links, S0xx supersedes/lifecycle, P0xx parser). Use when the user asks to lint ADRs, check ADR integrity, validate the corpus, or mentions "adr-fmt --lint".
Print the ADR domain tree (parent → child via AFM-0020 parent-edge model) optionally filtered by domain prefix, via `adr-fmt --tree [DOMAIN]`. Use when the user wants to enumerate ADRs in a domain, see the architectural hierarchy, or mentions "adr-fmt --tree".