Overview
The CDR index is injected into the system prompt at session start by team-boot. The LLM natively matches CDR descriptors against the current task and reads full module bodies on demand. This native matching happens per prompt without invoking this skill.
Use /team-discover when you want a structured discovery table with
explicit relevance assessments — for example, when starting a complex
feature or verifying which directives apply.
Relationship to team-boot
team-boot is the session bootstrap: it injects the CDR index into the
system prompt via the session_start event hook. Per-prompt discovery
matching is native — the LLM matches the injected CDR index against the
current message on its own, so team-boot does not invoke this skill
as part of its bootstrap loop.
/team-discover is user-invoked: run it when you want an explicit,
structured re-scan with a relevance-rated table (complex features, audits, or
verifying which directives apply). It is not required for normal operation.
Process
- Extract domain, technology, patterns, and actions from the task.
- Scan the CDR index for matching entries.
- Assess relevance: High / Medium / Low.
- Read full module bodies for High-relevance entries.
- Output:
| ID | Module | Type | Descriptor | Relevance |
|---|---|---|---|---|
| CDR-2026-003 | context_modules/personas/admin.md | Persona | Admin persona | High |
- Include PDR/ADR/ChDR matches from project indexes if relevant.
- Include:
_Searched N CDRs, M PDRs, K ADRs, C ChDRs, J matches found._
Absent Context
If no CDR index was injected at session start: report Searched 0 CDRs — index not injected (unconfigured project or hook failure), reading workspace memory directly, proceed against docs/adlc/memory/ on disk, and continue the task. Never present on-disk fallback rows as injected context. Recovery: run /team-diagnose.
Unconfigured projects
If team AI directives are not configured, invoke team-setup.