v1v2
role
trust
Craft
Maya Patel — craft posture
21of 27 checks passing
90-day checks-passing trend: 17 → 2190 days: 1721· top quartile (org, k≥5)
quality streak
12 sessions

sessions at/under your median tokens/accepted-diff — this run is 10,250,000 tokens under

PlanBefore generation: was the task framed, scoped, and plan-moded when it should be?
Plan-mode discipline
3/4PROXY

skipped on 6 multi-file sessions → ~2.1 avoidable rework loops ≈ 35,000,000 tokens/mo

Prompt quality
3/4PROXY

2 sessions/wk need 3+ clarifying loops — add done-conditions

Build · SuperviseDuring generation: right-sized models, warm cache, converging iterations instead of thrash.
Waste patterns
4/4MEASURED

top quartile — your cache reuse saves ~15,500,000 tokens/mo vs cohort median

Iteration efficiency
2/4MEASURED

9-loop session on auth spike — plan-mode restart suggested

VerifyBetween diff and accept: tests, typechecks, read-backs — accept-without-checking is the fast-and-reckless failure mode.
Edge-case handling
3/3MEASURED

edge-case test diffs added before the fix in 2 sessions — MEASURED from diff telemetry

Verification discipline
3/4MEASURED

1 accepted diff this week had no verify event between diff and accept

Correct · ReviewAfter accept: do accepted diffs stick, or get reworked shortly after?
Execution quality
3/4MEASURED

1 accepted diff re-edited in later sessions on the same files within 3 days

The same loop graded at hiring — measured continuously, on real work. Faking a signal costs real work to fake and leaves detectable signatures.

Impact Log and summaries are private to each engineer at your altitude.

what your manager sees · Accountability-focused

Accountability-focused: your named card — checks, token usage, and sessions — is visible to your manager.

Invariant in every mode: you always see more of your own data than anyone above you.

Compiled 07:30 from 1,204 sessions · quota weight table v3 (May 12) · measured vs proxy tagged · Accountability-focused