Frequency of per-session cost in $5 buckets, across every session (each (session, project) snapshot, not just /implement or issue-attributed work). Long-tailed — the median and quartiles describe the typical session better than the mean.
| Issue | Cost | Sessions |
|---|---|---|
| #122 | $54.23 | 2 |
| #142 | $26.88 | 3 |
| #71 | $26.07 | 1 |
| #138 | $22.64 | 1 |
| #111 | $21.53 | 1 |
| #256 | $20.91 | 1 |
| #85 | $15.24 | 1 |
| #422 | $13.88 | 1 |
| #177 | $11.48 | 1 |
| #86 | $11.21 | 1 |
Issue attribution covers 5.4% of session-level spend — only sessions in worktrees whose name embeds an issue number are counted. Operator-driven work in main and slug-named worktrees is not attributed.
Weekly outer-loop allocation & throughput (OUTER_LOOP.md §1, §10), one point per ISO week from .claude-team/process-metrics/. Queue and open-PR series begin where live snapshots start; earlier weeks are backfilled allocation only.
One point per ISO week from .claude-team/outer-loop/skill-runs.jsonl — one line per interactive LLM-skill tick (/elaborator, /po-todo, /implement), the sibling of the janitor run log. Series begin where logging started.
| Package | Lines | Branches | Functions | Statements | |
|---|---|---|---|---|---|
| ai | 97.1% | 89.6% | 98.3% | 97.3% | Details |
| shared | 100.0% | 100.0% | 100.0% | 100.0% | Details |
| web | 62.8% | 64.2% | 44.6% | 60.4% | Details |
Solid = ai, Dashed = shared, Dotted = web — Lines / Branches / Functions