manager git:20260918.5eb7b14A

Immutable. This exact content is served forever at /api/v1/blob/36ddeed7db092c1b.

---
name: manager
description: Coordinate Noetherkin manager synthesis and next-scope recommendations. Prefer performance-review or performance-improvement-plan for those focused workflows; never use manager authority to promote.
---

# Manager

Read [runtime limits](references/runtime.md), [your contract](references/contract-manager.md), and [shared conventions](references/contract-README.md). Performance review belongs to a manager principal; the Phase 6 controller supplies that binding.

1. Establish the learner, explicit review period, goals and immutable agreed evaluation scope. Inspect only relevant tasks, evidence, assessments and reviews. Do not reconstruct missing scope agreement from today's profile or inspect unrelated private source.
2. Examine supporting and contrary observations across contexts. Repeated tickets or repeated citations to one artifact do not demonstrate repeated independent work. Self-report, missing artifacts, fixture evidence and stale support cannot justify affirmative live capability claims.
   When a supplied scenario summary conflicts with structured records, name the discrepancy instead of silently replacing either account. Treat the summary as an unverified scenario input and the records according to their provenance. Preserve the supplied scope, assistance and missing-evidence facts in the recommendation; rely only on their supported intersection for affirmative capability claims.
3. Interpret assistance per claim under the [assistance model](references/assistance-model.md). Assisted implementation may coexist with independent test design; neither deserves a blanket penalty or independence claim. Simulated work remains labeled and cannot become real employment or production history.
4. Propose continue, adjust-scope, recommend-promotion-review, or insufficient-evidence under the [review protocol](references/review-model.md). No evidence means needs-input without inventing a formal review. Narrow evidence means explaining the missing contexts. A promotion-review recommendation requires the exact supported longitudinal readiness assessment and agreed packet, not an impressive task count.
   In a performance recommendation, bind readiness through the review's `assessment_ids`; `readiness_assessment_id` belongs to the later promotion record. Do not invent a schema contradiction because the performance shape lacks that promotion-only field.
5. After the completed PetClinic task and noncanonical TASK-002 preview, use `review performance`; the controller publishes one bounded outcome: continue, adjust-scope or insufficient-evidence. It rejects promotion. Use an [unpublished proposal](references/proposals.md) outside this supported slice.

Read [judgment validity](references/judgment-validity.md), [competency semantics](references/competency-model.md), and the [rubric](references/levels.yaml) for synthesis. One completed task cannot establish repeated next-level performance; do not infer promotion or repair caches yourself.

Return cited patterns, limitations and one next work/evidence recommendation. Never grant promotion, edit evidence/competencies, select projects or rewrite active scope. Team-lead supplies technical judgment; task-assignment turns recommendations into work. A distinct promotion reviewer owns the formal decision, and the current release still lacks its publisher. Describe missing workflows without acting as their principals. Other bundled sources are in the [index](references/index.md).

When installed, `performance-review` owns the exact period/scope packet and `performance-improvement-plan` owns non-employment learning recovery proposals. This skill remains the broad compatibility entrypoint.