Team Levelup

team-levelup

What this skill does

Session-end CDR lifecycle — extracts reusable patterns from the completed session, scores them by confidence, presents for batch review, and publishes accepted CDRs as a draft PR to team-ai-directives.

Replaces the former levelup-specify, levelup-clarify, and levelup-publish skills with a single streamlined workflow.

Searched N CDRs, K matched. 0 rows matched → emit the section heading + the searched line only — no table. A 0-row table header collapses into unrendered single-line markdown; never emit one.

When no local context-module records exist in the current working directory and the directory sits inside a workspace (detected via a .gitmodules marker in an ancestor), read the workspace root's records instead.

When to use

Invoke at the START of a matching task — before planning the todo list and before implementation — so the CDR context informs planning. Never defer to session end.

• Session end (automatic): .events.json maps session_end → this skill
• Manual invocation: /team-levelup after completing work
• Before closing a branch: Extract team-wide learnings

When NOT to use

• Brownfield discovery: Use /team-init to scan existing code
• ADR/PDR/ChDR capture: Use the respective clarify skills (architect-clarify, product-clarify, change-clarify)

Storage

CDR drafts live in the adlc orphan branch of team-ai-directives:

team-ai-directives (adlc branch, orphan)
├── drafts/cdr/
│   ├── CDR-001.md        # Pending CDR
│   └── cdr.md            # Draft index
└── reports/
    ├── sessions/<user>/<YYYY-MM>.md
    ├── projects/<project>.json
    └── confidence-scores.json

Main branch has NO drafts directory — only accepted CDRs in context_modules/.

Process

Phase 0: Environment Setup

Run the setup script:

scripts/team-levelup.sh --setup

Parse JSON for REPO_ROOT, TEAM_AI_DIRECTIVES, NEXT_CDR, ADLC_BRANCH_EXISTS.

If TEAM_AI_DIRECTIVES is not configured, exit with message to run /team-setup.

Phase 1: Extract CDRs from Session

Review the current session to identify reusable patterns:

1. What did the user ask for?
2. What did the agent do? (file changes, key decisions, approach)
3. What reusable patterns emerged?
4. What files were created/modified? (git diff --stat, git log --oneline -10)

For each pattern, create a CDR draft using the shared template at skills/team/templates/cdr-draft-template.md.

Write session trace to adlc branch: reports/sessions/<user>/<YYYY-MM>.md.

Phase 2: Score Confidence

For each extracted CDR, calculate confidence:

Base scores by type:

• Rule: 0.60
• Persona: 0.50
• Example: 0.50
• Constitution: 0.70

Bonuses:

• +0.20 if paired eval exists
• +0.10 if evidence includes file paths/commits
• +0.10 if multiple projects reference same pattern

Usage multiplier (from reports/confidence-scores.json):

• success_rate > 0.8: ×1.2
• success_rate 0.5-0.8: ×1.0
• success_rate < 0.5: ×0.8

Thresholds:

• ≥ 0.8: HIGH — batch review, auto-accept after review
• 0.5-0.79: MEDIUM — batch review required
• < 0.5: LOW — keep as draft

Phase 3: Batch Review

Present CDRs one at a time (same as former team-learn logic):

## CDR-{ID}: {Title}

**Context Type**: {type}
**Confidence**: {score} ({HIGH/MEDIUM/LOW})
**Current Status**: {status}

### Current Content
...

### Choose Action

| Option | Action |
|---|---|
| A | Accept — Approve for implementation |
| B | Reject — Decline with reason |
| C | Defer — Skip for now, keep pending |
| D | Accept all remaining — Accept this and all pending CDRs |
| P | Promote to check (mechanical rules only) |

Reply with your choice (A/B/C/D/P).

Wait for user input before proceeding. Update CDR file after each decision.

Absent Context

If team-boot injected no team context this session (no Team Context & Decisions section in the first user message — unconfigured project or hook failure): say so in one line, maintain rows in Team Context & Decisions with a 0 CDRs matched (no team context injected — run /team-setup) source line, and continue the task on directly-read CDR modules. Never treat a missing injection as an empty record set. Recovery: run /team-diagnose.