The key contract for the discovery plugin. Every key is optional: with nothing set, each skill
runs on the defaults below.
Resolved lowest first; a later layer wins (ADR 0064 Decision 8):
| Order | Layer | Where |
|---|---|---|
| 0 | default | the Default column below |
| 1 | per user | the plugin's userConfig option of the same name |
| 2 | repository | docs/conventions/discovery.yaml at the repository root, tracked, validated by schemas/discovery.schema.json |
/discovery:setup apply writes the repository file after validating it against the schema, and
shows the diff and asks before it changes an existing one. userConfig is the per-user layer;
no ~/.claude file, .claude/ file or personal override sets these keys. The skill reads the repository file itself; a missing file or key leaves that layer unset,
and the layer does not apply outside a git working tree. Every run that reads a key reports the value and
the layer that supplied it.
A value outside the key's values in any layer is reported, naming the layer's file (or
userConfig), the key and the value, and that layer is dropped: a valid higher layer still wins,
and with none the key resolves to its default; a lower layer's value is never used in its place. A userConfig value equal to the default is
reported as the default level, since the two cannot be told apart, and a literal unexpanded
${user_config.<key>} means the layer is unset.
| Key | Values | Default | Reader and effect |
|---|---|---|---|
explore_output |
auto, change-prep, explain |
auto |
/discovery:explore. Picks the final reply only; EXPLORE.md and its sidecars are written the same way under every value. change-prep replies with the handoff summary for the next stage (research or planning). explain replies with a walkthrough of how the scope works, each step citing path:line. auto picks explain only when a person asks how something works, and change-prep for anything else, including every call from another skill. A caller passes --output explain or --output change-prep to set it for one run; that argument wins over every layer. |
research_corroborators |
whole number, 1 or more | 2 |
/discovery:research. Independent current corroborators each accepted claim needs beside its one primary (discipline 5, outcome-gate row 4); the parent passes the resolved value to the researcher and the verifier. The base floor only: discipline 8's broad-topic minimum (4+ Tier 0/1) and the "Single-source first-party content claims" exception are unchanged, each claim keeps exactly one primary (scripts/check-source-applicability.py R5), and HIGH still needs three current Tier 0/1 pools (scripts/check-claim-eligibility.py E5). |
research_recency_active_days |
whole number, 1 or more | 14 |
/discovery:research recency gate: how stale a cited doc on a very active project may be before the changelog cross-check is required. |
research_recency_standard_days |
whole number, 1 or more | 30 |
/discovery:research recency gate: the same window for a standard library, tool, CLI or API. |
research_recency_guide_days |
whole number, 1 or more | 90 |
/discovery:research recency gate: the same window for a tool- or vendor-specific architecture guide. |
verification, verification_skills.<skill> |
off, light, standard, strict |
light |
/discovery:research, /discovery:explore and /discovery:trace-intent: the sibling verifier, through lib/read-verification.sh. Repository file only, read from the default branch; no user option; a skill entry beats the plugin value (verification level). |